Review of Accounting and Finance: Article Information

Review of Accounting and Finance
Assessing earnings management flexibility

Hiu Lam Choy,
Article information:
To cite this document:
Hiu Lam Choy, (2012) "Assessing earnings management flexibility", Review of Accounting and Finance,
Vol. 11 Issue: 4, pp.340-376, https://doi.org/10.1108/14757701211279169
Downloaded by UNIVERSITAS TRISAKTI, User Trisakti At 10:33 10 October 2018 (PT)
Permanent link to this document:

https://doi.org/10.1108/14757701211279169
Downloaded on: 10 October 2018, At: 10:33 (PT)
References: this document contains references to 47 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 3636 times since 2012*
Users who downloaded this article also downloaded:
(2015),"Managerial tenure and earnings management", International Journal of Accounting
& Information Management, Vol. 23 Iss 1 pp. 42-59 <a href="https://doi.org/10.1108/
IJAIM-04-2014-0025">https://doi.org/10.1108/IJAIM-04-2014-0025</a>
(2016),"Beating threshold targets with earnings management", Review of Accounting and Finance,
Vol. 15 Iss 2 pp. 198-221 <a href="https://doi.org/10.1108/RAF-04-2015-0057">https://doi.org/10.1108/
RAF-04-2015-0057</a>
Access to this document was granted through an Emerald subscription provided by emerald-srm:551360 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.
*Related content and download information correct at time of download.

The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1475-7702.htm
RAF
11,4 Assessing earnings
management flexibility
Hiu Lam Choy
340 Department of Accounting, Drexel University, Philadelphia, Pennsylvania, USA
Abstract
Purpose – The purpose of this paper is to propose a new measure of earnings management flexibility
based on the limits of the allowable set of accruals, prior discretionary accruals used, and the reversal
rate of these accruals.

Design/methodology/approach – Quarterly financial data from Compustat for the period
1990-2009 were used to construct the flexibility measure. Then the author examined how well this
measure captures flexibility by investigating its effect on a firm’s probability of meeting analysts’
forecasts.
Findings – The results show that this flexibility measure better captures the firm-specific flexibility
than that of Barton and Simko which captures mainly the difference in flexibility across industries.
Further, the positive effect of their measure on a firm’s probability of meeting/beating analysts’
forecasts is not observed in the extended sample period.
Practical implications – The flexibility measure proposed here can assist investors, analysts, or
researchers to compare earnings management flexibility across firms in the same industry, which is
useful in evaluating the quality of a firm’s financial reports, stock picking or credit granting decisions.
Originality/value – This paper contributes to the earnings management literature by incorporating
both the variation in flexibility used and that in flexibility limits. Second, evidence in this paper
suggests that while financial benefits motivate managers to undertake earnings management,
flexibility determines the extent of earnings management they can undertake. Third, this study points
out that the unreversed discretionary accruals impose a constraint on the level of discretionary
accruals a manager can incur in the current period, and hence have an indirect influence on current
reported earnings.
Keywords Earnings, Financial flexibility, Stock markets, Financial markets, Earnings management,
Flexibility, Accruals, Analysts’ forecasts
Paper type Research paper
1. Introduction
This paper proposes a new construct of earnings management flexibility based on the
variation in both the earnings management flexibility limit and the flexibility used. Prior
earnings management studies focus on identifying managers’ incentives to manipulate
earnings (Healy, 1985; Sweeney, 1994; DeFond and Jiambalvo, 1994; Burgstahler and
JEL classification – M40, M41

This paper is based on the author’s dissertation at University of Rochester. The author would
like to thank her committee members Andy Leone, Ross Watts, and especially her chair
Jerold Zimmerman, for their advice and comments. The author appreciates helpful comments from
Janie Chang (editor), two anonymous reviewers, Pam Bedora, Hsihui Chang, Elizabeth Demers,
Review of Accounting and Finance
Vol. 11 No. 4, 2012 Assaf Eisdorfer, Philip Joos, Yuanzhi Luo, Claudine Mangen, Sangwoo Shin, Sabatino Silveri,
pp. 340-376 Mark Vargus, Charles Wasley, and participants at the workshops of Boston University,
q Emerald Group Publishing Limited
1475-7702
CUNY Baruch, INSEAD, Pennsylvania State University, SUNY Buffalo, University of California
DOI 10.1108/14757701211279169 at Riverside, and University of Texas at Dallas.
Dichev, 1997; Teoh et al., 1998a, b; Erickson and Wang, 1999; Degeorge et al., 1999; Earnings
Fields et al., 2001). While these studies usually show that earnings management exists, management
there is a missing link: the manager’s latitude to manipulate earnings (Fields et al., 2001).
Accounting conventions of objectivity and verifiability limit the set of allowable flexibility
accruals (Watts and Zimmerman, 1986; Dechow, 1994). GAAP also constrains the
manager’s flexibility in managing earnings. Both internal and external control
mechanisms, such as auditors, outside directors, audit committees, and regulators, help 341
constrain managers’ discretion. Although managers can incur accruals beyond the
allowable set (e.g. by committing fraud), the costs of violation[1] are high and likely
outweigh its benefits for most firms. In most situations, managers will stay within the
allowable set of accruals[2] which I denote as the “earnings management flexibility

limits.” Given the limit on the accruals, the earnings management flexibility available to
the manager at any specific time equals the limit minus the accruals already taken in
prior periods.
Two papers – Barton and Simko (2002) and Kasznik (1999) – examine the effect of
earnings management flexibility. Using the firm’s net operating assets (i.e. shareholders’
equity less cash and marketable securities, plus debt) relative to sales as a proxy for
earnings management flexibility, Barton and Simko (2002) suggest that firms with low
flexibility have difficulty in managing earnings up by even 1 cent per share to meet the
analysts’ forecasts. Kasznik (1999) finds that firms with low flexibility (proxied by the
change in total accruals in the preceding year) have difficulty in meeting their own
management forecasts. The underlying idea of both papers – a firm’s prior earnings
management affects current earnings management flexibility – is appealing. This paper
adds to the literature by analyzing the determinants of earnings management flexibility,
discussing the three different effects prior earnings management practices have on a
firm’s ability to manage earnings, and proposing a flexibility construct that incorporates
all these factors. Empirical tests show that this new flexibility construct complements
the flexibility measures proposed in prior studies and better captures the firm-specific
component of earnings management flexibility.
In this paper, a measure of flexibility available (operating cycle flexibility measure) is
constructed. This flexibility measure is based on the difference between flexibility limit
and flexibility used. Barton and Simko’s (2002) and Kasznik’s (1999) measures capture
mainly the flexibility consumed in prior periods. They implicitly assume that the set of
allowable accruals is constant across firms and over time. I incorporate the variation in
flexibility limit in my measure because the set of allowable accruals, and hence
flexibility limits, vary both cross-sectionally and over time with a firm’s operating
characteristics, growth, governance structure, quality of auditors, and the regulatory
environment. A firm’s remaining flexibility depends on both its past earnings
management practice and its flexibility limits. It is the flexibility available, rather than
just the portion consumed, that will affect the manager’s earnings management decision.
In constructing the proxy for flexibility used, I account for both the earnings
management practices in prior periods and the reversal rate of accruals. Whereas prior
period earnings management practices determine the level of flexibility used in one
period, the reversal rate determines the number of subsequent quarters that are
affected by these earnings management practices. I incorporate the variation in the
reversal rate of accruals, proxied by various measures reflecting a firm’s operating
characteristics, in my flexibility construct.
RAF To compute the level of earnings management in prior periods, I use the cross-sectional
11,4 Jones model to estimate discretionary accruals, my proxy for earnings management. This
measure accounts for the systematic difference in operating characteristics across
industries and specific firm effects (DeFond, 2002), such as an increase in sales in the
current period, which are not accounted for in Barton and Simko’s (2002) measure.
I compare the operating cycle flexibility measures with Barton and Simko (2002)
342 and Kasznik (1999) measures by examining the measures’ effect on a firm’s probability
of missing analysts’ forecasts. Within the Barton and Simko (2002) sample period
(1993-1999), both my operating cycle flexibility measure and Barton and Simko’s
measure have a significant impact on the probability of missing analysts’ forecasts.
However, Kasznik’s measure does not have any significant effect. This result suggests
that both my measure and Barton and Simko’s measure capture a firm’s flexibility to
manage earnings. In an extended sample period (1990-2009), the only flexibility
measure that has a significant impact on the probability of meeting analysts’ forecasts
is my cross-sectional operating cycle flexibility measure. Further, this result holds in
both pre- and post-SOX periods, and after controlling for lagged performance. When
I include my flexibility measure and either Barton and Simko’s measure or Kasznik’s
measure in the model simultaneously, only my measure has a significant association
with a firm’s probability of meeting analysts’ forecasts.
Additional tests show that when industry-adjusted flexibility measures are used,
Barton and Simko’s flexibility measure no longer has a significant influence on the
probability of missing analysts’ forecasts even in their sample period. Only the
industry-adjusted cross-sectional operating cycle flexibility measure shows a
significantly negative correlation. Taken together, these results suggest that while
Barton and Simko’s measure captures the operating characteristics of different
industries (or the industry component of flexibility), my measure further captures the
firm-specific component of flexibility. In addition, the flexibility measures suggested in
prior studies are sensitive to the sample period examined. This variation in results can
be caused by the underlying assumption of their measure – the limit on flexibility is the
same across firms and over time. With changes in the regulation and corporate
governance, the limit on earnings management flexibility likely changes as well. My
flexibility measure accounts for the impact of these changes by considering both the
flexibility limit and the flexibility used in each period.
This paper contributes to the extant literature on earnings management in three
ways. First, my measure of earnings management flexibility incorporates the variation
in flexibility used and the flexibility limits. I demonstrate that my flexibility measure
provides incremental information to both Barton and Simko’s and Kasznik’s flexibility
measures. My measure captures the difference in earnings management flexibility
across firms within the same industry rather than just the difference across industries.
Furthermore, accounting for the difference in flexibility limits is important for studies
examining an extended sample period because flexibility limits could vary with
changes in the legal and business environment.
Second, evidence in this paper suggests that while financial benefits motivate
managers to undertake earnings management, flexibility determines the extent of
possible earnings management. Using both the flexibility measure and the managers’
incentives can lead to further insights into the variation in earnings management both
across firms and over time.
Third, this study supplements existing research on earnings management by Earnings
pointing out that the unreversed discretionary accruals impose a constraint on the level management
of discretionary accruals a manager can incur in the current period, and hence have an
indirect influence on current reported earnings. Accordingly, both the reversal of flexibility
discretionary accruals and the unreversed accruals affect the earnings management
flexibility. Further, the reversal rate of accruals becomes a significant determinant of a
firm’s flexibility because it determines the number of periods affected by the 343
unreversed discretionary accruals.
The rest of the paper is organized as follows. In Section 2, I analyze current earnings
management flexibility. Section 3 discusses proxies for earnings management
flexibility. Section 4 describes other factors affecting a firm’s flexibility and their
proxies. Section 5 describes sample selection and descriptive statistics for the sample.
In Section 6, I compare my flexibility measure with Barton and Simko (2002) and Kasznik
(1999) measures. Section 7 describes the sensitivity tests and Section 8 concludes.
2. Effect of prior earnings management and reversal rate on current

earnings management flexibility
In this section, I discuss the three determinants of a firm’s earnings management
flexibility: prior earnings management practices, reversal rate of accruals, and
flexibility limits. I first analyze the various effects prior earnings management
practices have on current quarter flexibility. Then I describe the effect of the reversal
rate of accruals and that of flexibility limits on the current flexibility available.
Effect of prior earnings management practice

The following analysis is based on the intertemporal relation between accruals, and the
assumption that earnings management is not free. Based on these assumptions, I show
that prior earnings management affects current costs of earnings management and
flexibility available in three different ways:
(i) Reversal effect. Since total discretionary accruals have to sum to zero over a firm’s
life, inflated earnings in one quarter must reverse in subsequent quarters. These
reversals lower the reported earnings of the subsequent quarters (Dechow, 1994;
Barton and Simko, 2002; Kasznik, 1999; McCulloch, 1998). For instance, in 2002
Amdocs Limited was sued for understating reserves for doubtful accounts in fiscal
2001. When the accounts became uncollectible in 2002, the firm had to write-off these
accounts. These write-offs caused its earnings to drop significantly in 2002 and
triggered the lawsuits (Business Wire, 2002).
(ii) Constraints effect. A firm’s net operating assets and accruals are closely tied to
its operating activities. Firms with abnormally high levels of accruals relative to their
operating activities are seen either as inefficient in utilizing assets or as overstating
their values, which will arouse the suspicion of analysts, auditors and regulators.
Therefore, incurring accruals beyond the allowable set is costly.
For example, the accounts receivable balance can be considered as the cumulative
change in accounts receivable since the founding of the firm: the initial incurrence of
accounts receivable minus the collection and write-offs. Assuming the net accounts
receivable balance is bounded between K and K:

K # B_Accrut þ Accrut # K: ð1Þ
RAF B_Accrut is the unreversed accruals carried forward from last period and Accrut is the
11,4 level of accruals incurred in quarter t (i.e. the change in net accounts receivable).
The higher is the level of unreversed positive accruals, B_Accrut, the lower the level of
Accrut that can be incurred in the current quarter[3]. Unlike the reversal effect of
positive accruals, which always lowers the current reported earnings, the unreversed
positive discretionary accruals constrain a firm’s flexibility only if its current
344 pre-managed earnings are below certain targets (such as analysts’ consensus forecasts)
and the firm has the need to manage earnings. That is, the unreversed discretionary
accruals constrain the accruals a firm can incur.
(iii) Ratchet effect. In addition to the reversal and constraint effects, earnings
management in the current quarter raises analysts’ and investors’ expectation of future
earnings and the performance benchmark of future quarters (Leone and Rock, 2002;
Murphy, 2000). Given that the pre-managed earnings in two quarters are the same,
inflating earnings by $X in the first quarter can cause an “apparent” drop in the second
quarter if the manager does not inflate earnings in the second quarter by at least $X.
This again raises the costs of earnings management. Ceteris paribus, this leads to a
higher probability of missing the benchmark in subsequent quarters.
The above analysis implies that accruals, both discretionary and nondiscretionary,
incurred in prior quarters affect reported earnings and flexibility in the current quarter.
However, since nondiscretionary accruals are usually a function of the firm’s operating
activities and the market’s expected level of accruals incorporates the reversal effect of
these accruals, current earnings management flexibility is constrained only by
discretionary accruals in prior quarters. Consequently, I concentrate on the effect of prior
quarters’ discretionary accruals on earnings management flexibility in the following
analysis.
Effect of the reversal rate of accruals

The reversal rate of accruals determines the number of subsequent quarters affected
by the above three effects of accruals. A slower reversal rate causes the current
accruals to reverse in a later quarter and hence affects the reported earnings of that
quarter. The slower rate also increases the number of quarters whose flexibility limits
are lowered by current accruals (i.e. those periods between the current quarter and the
quarter of reversal), and hence, decreases the manager’s discretion in those quarters.
Consequently, a firm’s flexibility in any quarter is a function of both its prior
discretionary accruals and the reversal rate of these accruals.
Effect of the width of flexibility boundaries

In addition to earnings management in prior quarters, the flexibility limits (i.e. K; K )
constrain the current flexibility available. These limits represent the points beyond
which costs of earnings management are prohibitively high. The nature of the firm’s
business and its operating activities determine the set of GAAP applicable to the firm.
The governance structure and both the internal and external monitoring mechanisms
affect the choice of accounting methods from the menu of acceptable methods set by
GAAP and the probability of being detected when the rules are violated. All of the
above factors determine the allowable set of accruals and set the flexibility limits of a
firm. Given the discretionary accruals incurred in prior quarters and the reversal rate
of accruals, the tighter are a firm’s flexibility limits, the smaller is the flexibility Earnings
available to the manager in the current quarter. management
3. Empirical proxies for flexibility
flexibility
The above analysis argues that a firm’s available flexibility depends on the flexibility
limits, the discretionary accruals incurred in prior quarters, and the reversal rate of those
accruals[4]. Specifically, the available flexibility of a firm-quarter is the difference 345
between the flexibility limits and the cumulative discretionary accruals not yet reversed
at the beginning of the quarter. In this section, I discuss the construction of empirical
proxies for discretionary accruals, reversal rate of accruals, flexibility limits, and
flexibility available. While prior discretionary accruals theoretically have three different
effects on current flexibility, it is difficult to distinguish the reversal effect from the
constraint effect. Consequently, my flexibility proxy captures these two effects together
as the impact of prior earnings management. The ratchet effect is captured by the level of
the benchmark against which current performance is evaluated.
Discretionary accruals
I use the cross-sectional Jones model adjusted for growth to estimate the level of
discretionary (unexpected) accruals (a proxy for earnings management) incurred in the
prior quarters. The cross-sectional Jones model assumes that all firms in the same
industry and same quarter have the same non-discretionary (expected) accruals/Dsales
ratio. However, the level of non-discretionary accruals relative to the change in sales
likely depends on the firm’s stage in its life-cycle. Growth firms likely have a higher
ratio of expected accruals to sales change than mature firms to support their growth.
In order to account for these different growth rates across firms within the same
industry, I include two of Anthony and Ramesh’s (1992) proxies for a firm’s life-cycle
stage – average sales growth (SG) and average dividend payout ratio (Divid ) in the
prior five years – in the cross-sectional Jones estimation model:
TAt a DSalest PPE t

¼ þ b1 þ b2 þ b3 SGt þ b4 Divid t þ 1t :
TAsset t21 TAsset t21 TAsset t21 TAsset t21
ð2Þ
where:
TAt ¼ total accruals of the firm-quarter, computed as net income before
extraordinary items less net operating cash flow.
DSales ¼ difference between sales of this quarter and that of the same quarter last
year.
PPE ¼ average gross property, plant, and equipment of the quarter.
These three variables (TA, DSales, PPE) and the intercept term (a) are all scaled by
total assets measured at the beginning of the quarter. As the above model adjusts for
the industry and firm characteristics in calculating the expected accruals, the
discretionary accruals thus calculated provide a more precise measure of the degree of
earnings management in prior periods (DeFond, 2002).
RAF Reversal rate
11,4 I use a firm’s operating cycle as a proxy for the reversal rate of accruals because it
reflects the time needed to recover the firm’s operating costs. Operating cycle captures
the average number of days between ordering (and paying) for the raw
materials/inventories and selling the inventories (and collecting the money from
customers). It provides an estimate for the average time it takes the working capital
346 accruals to reverse. The reversal rate of the discretionary working capital, in turn,
estimates the reversal rate of the discretionary accruals because a large part of the
discretionary accruals are from the working capital accounts[5]. A firm’s operating
cycle is computed as:

Avg A=R Avg Inventory Avg A=P

Operating cycle ¼ þ 2 * 90 ð3Þ
Sales Cost of goods sold Cost of goods Sold
I compute the average operating cycle both cross-sectionally (CS operating cycle) and
over time (TS operating cycle). To compute CS operating cycle, firms in the same
two-digit SIC codes and the same quarter are grouped together in order to calculate the
average operating cycle of that industry based on equation (3). This average operating
cycle is then used as a proxy for the reversal rate of all firms in that industry-quarter.
The TS operating cycle is computed as the average operating cycle of the firm itself
(the operating cycle is computed as in equation (3)) in the prior 12 quarters[6]. Results
using the TS operating cycle reversal rate is similar to those using the CS operating
cycle reversal rate. Thus, only results of flexibility measure based on the CS operating
cycle reversal rate are reported in the following sections[7].
Proxy for earnings management flexibility used

After estimating discretionary accruals and reversal rates, I calculate the cumulative
lagged discretionary accruals, a proxy for flexibility used in prior quarter and not yet
reversed at the beginning of the period, for each quarter in the sample period[8]. If the
cumulative lagged discretionary accruals are positive, they constrain the discretionary
accruals a firm can incur in the current quarter because of either the reversal or the
constraint effect. On the other hand, if the cumulative lagged discretionary accruals are
negative, it provides the firm with extra earnings management flexibility. This
additional flexibility comes from the cookie jar reserve the firm built up in prior
quarters. The number of lags included in the cumulative measure is based on the
reversal rate of each firm’s discretionary accruals. For example, if a firm’s accruals
reverse on average in three quarters, then its cumulative lagged discretionary accruals
at time t (CLDAt) equal the sum of discretionary accruals in the previous three
quarters. That is, CLDAt ¼ DAt21 þ DAt22 þ DAt23 .
Proxy for earnings management flexibility limits

Next, I compute a proxy for a firm’s flexibility limit. Although GAAP prescribes a
firm’s set of accounting rules, auditors, analysts, audit committee, and regulators
define what constitutes a reasonable interpretation of GAAP (Francis et al., 1999) and
monitor firms’ compliance with GAAP. That is, the monitoring by these parties
imposes a limit on a firm’s discretion in reporting accruals. One standard procedure in
auditing is to perform analytical analyses (such as ratio analysis with the ratios in
prior periods and industry average serving as the benchmark) to identify any potential
problem area (SAS 56). If a firm shows a significant increase in accruals in the current Earnings
quarter or has significantly larger accruals than its peers, this can potentially raise a management
red flag. Hence, I use the “normal” level of cumulative discretionary accruals as a proxy
for the flexibility limit. flexibility
To estimate the “normal” level of cumulative discretionary accruals, I first compute
the mean and standard deviation of the cumulative lagged discretionary accruals
(CLDA) of a firm using: 347
(1) its own CLDA of the previous three years; and
(2) the CLDA of all firms in the same industry-quarter.
As the tenor of the results does not vary whether these statistics are computed using
the time-series or cross-sectional data, only results based on the cross-sectional data
are reported in the following sections. Because CLDA are serially correlated, I compute
the Newey-West adjusted standard deviation. As CLDA proxies for the flexibility used
in a quarter, its range captures the limits of cumulative accruals a firm can have in
a quarter. To reduce the impact of outliers, I use the mean plus two standard deviations
of CLDA of all firms in the industry (instead of the range of CLDA in the industry) as a
proxy for the upper limit, Max_flex[9]. Assuming a normal distribution, these
estimates cover about 95 percent of the CLDA observations in the industry at time t.
The upward earnings management flexibility of a firm available at time t, Flex, is then
proxied by Flext ¼ Max_flext 2 CLDAt. Because the operating cycle is used as a
proxy for the reversal rate of discretionary accruals, I denote this earnings
management flexibility as the operating cycle flexibility.
To test whether this empirical proxy captures a firm’s earnings management
flexibility, I examine its effect on the probability of missing analysts’ forecasts in
Section 6. This is a joint test of the hypothesis that the constructed measure actually
captures a firm’s available flexibility and the hypothesis that, ceteris paribus, firms
with higher flexibility are less likely to miss analysts’ forecasts.
4. Other determinants of costs of earnings management and flexibility

In addition to prior earnings management, a firm’s current flexibility also depends on
other firm characteristics and macro-economic factors. In this section, I examine two
other determinants of flexibility – expected growth and the overall economic condition.
Expected growth
A firm’s life-cycle stage affects its earnings management flexibility. A high-growth firm
can justify an increase in net operating assets as a way to meet its future demand.
However, this increase in net operating assets also provides managers with more leeway
in shifting earnings across periods. As such, growth firms likely have more flexibility
than mature firms. Although I have controlled for a firm’s past growth in estimating
discretionary accruals in equation (2), a firm’s flexibility depends on investors’ expected
future growth of the firm. Thus, I include the lagged market-to-book of equity ratio, as a
proxy for the expected future growth of a firm, in the following tests.
Overall economy
Lev and Thiagarajan (1993) show that during economic booms, investors respond less
negatively to a disproportionate inventory increase (relative to increase in sales) than
RAF during recessions. When the economy is booming, an increase in operating assets can
11,4 actually signal a firm’s expansion plan to meet the expected increase in demand; when
the economy is in recession, however, investors expect a decline in sales and an
increase in bad debt write-offs. A disproportionately high level of inventory and
accounts receivables can signal a slow response to the deteriorating economy.
Similarly, in a high inflation environment, investors react more negatively to a build-up
348 of working capital (Lev and Thiagarajan, 1993) because the opportunity costs of
holding inventories and accounts receivable are higher. Thus, the upper flexibility
limit, and hence the ability to inflate earnings, is likely to be lower during a
recession/high inflation period. Following Lev and Thiagarajan (1993), inflation is
proxied by the annual percentage change in the Consumers’ Price Index and economic
growth is measured as the annual percentage change in real GDP.
5. Sample selection and data description

Sample selection
The empirical analysis is based on quarterly financial data collected from Compustat for
the period 1990-2009. The beginning of the sample period is determined by data
availability for computing flexibility limits. Assuming a reversal rate of three years,
I require at least three years of accruals data to compute the cumulative lagged
discretionary accruals for constructing the cross-sectional operating cycle flexibility
measure. Since the cash-flow statement data required to compute accruals are available
only after 1987, the sample starts from 1992[10]. I retrieve analysts’ forecast data from
the Institutional Brokers Estimate System (I/B/E/S), GDP data from the Bureau of
Economic Analysis web site, and CPI data from the Bureau of Labor Statistics web site.
Financial and regulated utilities sectors are excluded from the sample. Any
industry-quarter (industry is defined by the two-digit SIC codes) with less than ten firms
is also excluded because the estimation of the cross-sectional Jones model and flexibility
limit is likely to be imprecise. Any firm-quarter with an absolute value of scaled
discretionary accruals (scaled by lagged total assets) greater than one is also deleted.
The size of the cross-sectional operating cycle flexibility sample is 71,177 firm-quarters.
Sample description
Table I Panel A provides the descriptive statistics of the sample. The first three columns
present the summary statistics of the full sample. Sample firms have mean total assets of
$3.532 billion and median assets of approximately $0.612 billion. The average flexibility
available (computed as flexibility limit minus flexibility used), Op_flex, is 13.4 percent of
total assets. The mean flexibility limit (computed as the mean plus two standard
deviations of CLDA of the industry-quarter), Max_flex, is about 13.1 percent of total
assets, and the average flexibility used (CLDA) is 20.3 percent of total assets.
The negative average flexibility used and the negative mean discretionary accruals
of 20.2 percent suggests that firms, on average, are building up their cookie jar reserve of
discretionary accruals during my sample period. This reserve allows firms to take a larger
amount of discretionary accruals than normally allowed by their flexibility limits
in the future. The mean lagged net operating asset (Barton and Simko’s flexibility
measure) is about 4.628 times lagged sales while the average lagged change in total
accruals is 239.5 percent. Approximately 67.35 percent of firms are able to meet or beat
the analysts’ forecast.
(4) (5) (6) (7) (8) (9) (10) (11)

(2) Firms in the lowest Firms in the highest Lowest flexibility quintile-highest
(1) All firms (3) flexibility available quintile flexibility available quintile flexibility quintile
Mean Median SD Mean Median SD Mean Median SD t-test Wilcoxon z-statistic
Panel A: comparison of firm characteristics between firms in the lowest vs highest earnings management flexibility quintiles
Flexibility available
(Op_flex) 0.134 0.113 0.136 0.039 0.045 0.028 0.282 0.229 0.238 2120.81 * * * 2146.16 * * *
Flexibility limit
(Max_flex) 0.131 0.112 0.123 0.068 0.062 0.032 0.243 0.206 0.232 289.41 * * * 2142.25 * * *
Flexibility used
(CLDA) 20.003 20.001 0.054 0.029 0.017 0.044 20.039 20.020 0.089 81.24 * * * 103.05 * * *
Lagged NOA
(deflated by lagged
sales) 4.628 2.064 217.74 2.763 1.965 69.219 6.229 2.047 293.14 21.37 2.607 * * *
Lagged DTotal
accruals 20.395 20.424 5.607 20.410 20.360 5.212 20.317 20.443 5.870 21.42 3.439 * * *
Total assets at the
beginning of quarter
(in $ millions) 3,532.12 611.52 12,694.71 4,844.38 828.01 17,794.19 2,828.37 462.00 10,076.61 11.77 * * * 26.05 * * *
Lagged market-to-
book of equity ratio 3.455 2.289 92.438 2.241 2.143 41.078 5.083 2.311 191.894 21.73 * 28.55 * * *
Operating cycle (no.
of quarters) 1.857 2.00 1.083 1.606 1.000 1.023 1.983 2.00 1.251 227.86 * * * 238.95 * * *
Average growth in
sales in past five
years 0.307 0.121 7.913 0.318 0.108 8.903 0.323 0.135 6.279 20.05 217.31 * * *
Average dividend
payout ratio in past
five years 0.242 0 2.706 0.274 0.028 3.519 0.149 0 1.721 3.83 * * * 30.28 * * *
(continued)
management
Earnings
Descriptive statistics
flexibility
for sample firms

349
Table I.
11,4
350
RAF
Table I.
(4) (5) (6) (7) (8) (9) (10) (11)
Total accruals
(deflated by lagged
total assets) 20.014 20.012 0.054 20.016 20.013 0.053 20.015 20.013 0.069 21.28 22.46 * *
Discretionary
accruals
(DA_adjusted) 20.002 20.001 0.051 20.003 20.001 0.047 20.004 20.002 0.066 2.03 * * * 1.64 *
Pre-managed
Dearnings
(Pre_surp) 20.001 20.000 0.059 20.000 20.000 0.058 20.002 20.002 0.075 2.64 * * * 4.732 * * *
% firms meeting/
beating analysts’
forecasts 67.35 1 46.89 64.34 1.00 47.90 69.20 1.00 46.17 28.72 * * * 28.70 * * *
Forecast errors 20.007 0.008 0.594 20.013 0.006 0.821 20.009 0.010 0.517 20.41 25.89 * * *
ROA 0.008 0.014 0.053 0.014 0.015 0.045 20.006 0.010 0.079 26.46 * * * 25.77 * * *
Panel B: summary statistics of flexibility components across SIC sectors
Flexibility available Flexibility limit Flexibility used
SIC sector
Agriculture,
forestry, and fishing 0.103 0.086 0.086 0.091 0.086 0.068 20.012 20.004 0.039
Mining 0.112 0.078 0.200 0.112 0.078 0.200 0.001 0.001 0.033
Construction 0.072 0.056 0.085 0.071 0.063 0.075 20.001 0.001 0.035
Manufacturing 0.135 0.116 0.137 0.133 0.116 0.124 20.003 20.000 0.055
Wholesale trade 0.133 0.114 0.200 0.133 0.114 0.197 0.000 20.000 0.046
Retail trade 0.090 0.079 0.064 0.090 0.081 0.050 0.000 0.000 0.040
Service 0.165 0.143 0.124 0.158 0.136 0.102 20. 007 20.004 0.063
(continued)
(4) (5) (6) (7) (8) (9) (10) (11)

Definition of SIC sectors

Agriculture, SIC Mining SIC 1000-
forestry, and fishing 0100-0900 1400
Construction SIC 1500- Manufacturing SIC 2000-
1700 3900
Wholesale trade SIC 5000- Retail trade SIC 5200-
5100 5900
Services SIC 7000-
8800
Notes: Panel A: significant at: *10, * *5 and * * *1 percent levels (two-tailed); where, Op_flex – the cross-sectional operating cycle flexibility available, which is
calculated as the upper flexibility bound (Max_flex) less the sum of discretionary accruals cumulated over Qi quarters (CLDA), where Qi is the number of quarters in
the firm’s operating cycle; a firm’s operating cycle is proxied by the mean industry operating cycle, computed using the operating cycles of all firms in the same
quarter and industry; the discretionary accruals used to construct Op_flex are those estimated by:
TAt a DSalest PPE t
¼ þ b1 þ b2 þ b3 SGt þ b4 Dividt þ 1t ð2Þ
TAssett21 TAssett21 TAssett21 TAssett21
The upper flexibility bound is calculated as the mean of the industry’s (as defined by two-digit SIC) cumulative lagged discretionary accruals plus two times the
standard deviations of these cumulative lagged accruals; Max_flex – the upper flexibility bound, which is calculated as the mean of the industry’s (as defined by
two-digit SIC) cumulative lagged discretionary accruals plus two times the standard deviations of these cumulative lagged accruals; CLDA – flexibility used to
manage earnings in the prior quarters and have not yet reversed at the beginning of the current quarter; it is calculated as the sum of discretionary accruals
cumulated over Qi quarters, where Qi is the number of quarters in the firm’s operating cycle; Lagged NOA – lagged net operating assets deflated by sales; Net
operating assets are defined as shareholders’ equity less cash and marketable securities, plus debt; Lagged change in total accruals – the difference between the
total accruals of this quarter and those of the same quarter last year, deflated by total accruals of the same quarter last year; Operating cycle – operating cycle
estimated cross-sectionally for each industry-quarter; Industry is defined by the two-digit SIC code; the operating cycle is computed as the number of days in
inventory þ number of days in accounts receivable – number of days in accounts payable; that is:
(continued)
management
Earnings
flexibility
351
Table I.
11,4
352
RAF
Table I.

Operating cycle ¼ þ 2 * 90
Sales Cost of goods Sold Cost of goods Sold
Total accruals – net income before extraordinary items less net operating cash flows; DA_adjusted – discretionary accruals estimated using equation (2) in the
paper (also stated above in the definition of Op_flex); Pre_surp – the pre-managed earnings change of the current quarter deflated by lagged total assets; the pre-
managed earnings of a quarter are defined as earnings before extraordinary items and discontinued operations less discretionary accruals estimated by the cross-
sectional Jones model adjusted for sales growth and dividend payout ratio; the pre-managed earnings change is then calculated as the difference between the current
pre-managed earnings and the reported earnings of the same quarter last year [i.e. (REt 2 DAt 2 REt2 4 )]; % firms meeting/beating analysts’ forecasts – % of
sample firm-quarters that reported earnings equal to or above the last consensus analysts’ forecast observed before earnings announcement; forecast errors – the
difference between reported earnings and the last consensus analysts’ forecast observed before earnings announcement; ROA – income before extraordinary items
deflated by total assets at the beginning of the quarter; Panel B: this panel presents summary statistics on the components of our flexibility measure for firms in each
industry sector: flexibility available is presented in columns (1)-(3), flexibility limit is shown in columns (4)-(6), and flexibility used is displayed in columns (7)-(9):
Flexibility Available ¼ Flexibility Limit 2 Flexibility Used
Industry are defined by the SIC sectors; financial industries and regulated utility sectors are excluded from the sample; the grouping of firms into the different
sectors is provided at the end of this table
A comparison of firm-quarters in the lowest flexibility available quintile (columns Earnings
4-6 of Table I Panel A) with those in the highest quintile (columns 7-9 of Table I management
Panel A) shows that high-flexibility firms have both higher flexibility limits
(Max_flex) and lower flexibility used (CLDA) than low-flexibility ones. In fact, firms flexibility
in the highest flexibility quintile have an average CLDA of 2 3.9 percent of total
assets, suggesting that they are building up their reserves of accruals while firms
in the lowest quintile have incurred an average CLDA of 2.9 percent. The difference 353
in Barton and Simko’s measure (i.e. lagged NOA) and Kasznik’s measure (i.e. lagged
DTotal accruals) between the two groups is not statistically significant when
the t-test is used but is significantly positive when the Wilcoxon signed rank test
is used. Based on the Wilcoxon z-statistics, the cross-sectional operating cycle
flexibility measure is consistent with these two flexibility measures proposed in

prior studies.
Table I Panel B presents summary statistics on the flexibility measure and its
components (i.e. flexibility limits and flexibility used) for each industry sector. The
service sector has the largest flexibility available (0.165) and the highest flexibility
limit (0.158). The service sector has built up reserves in prior quarters, as reflected by
the 2 0.007 flexibility used. The mining industry, on the other hand, has used up the
most flexibility in the prior quarters (0.001).
6. Flexibility and missing analysts’ forecasts

In this section, I examine the relation between a firm’s upward earnings management
flexibility and its probability of missing analysts’ forecasts. Various studies document
that managers have strong incentives to meet/beat analysts’ forecasts. They also show
that the frequency of small positive forecast errors far exceeds that of small negative
ones – the middle asymmetry (Degeorge et al., 1999; Matsumoto, 2002). Abarbanell and
Lehavy (2003) further show that this middle asymmetry can be caused by earnings
management because it disappears when forecast errors are based on reported
earnings stripped of unexpected accruals. With strong incentives to manage earnings
to meet/beat analysts’ forecasts, the manager’s ability to do so, however, is constrained
by the flexibility available. This provides a great setting for examining whether my
measure provides any incremental information to the measures proposed in Barton and
Simko (2002) and Kasznik (1999). Specifically, I hypothesize that, ceteris paribus,
a firm’s probability of missing analysts’ forecasts decreases as its flexibility
increases. If the construct captures a firm’s earnings management flexibility, it should
have a significant negative effect on the firm’s probability of missing analysts’
forecasts.
I describe the probit model used to analyze the relation, provide descriptive
statistics on the control variables, and present the regression results in the following
subsections.
Empirical analysis
To examine the relation between earnings management flexibility and a firm’s probability
of missing analysts’ forecasts, I estimate the following probit regression model:
Pr ðmissi;t ¼ 1Þ ¼ Fða þ b1 Flexi;t þ b2 DDGDP t þ b3 DDCPI t þ b4 lpbi;t
ð4Þ
þ b5 Pre_surpi;t þ b6 Exp_sur i;t Þ
RAF The variable miss takes on a value of 1 if there is a negative earnings surprise (i.e. reported
11,4 earnings – consensus analysts’ forecast , 0) and zero otherwise[11], [12]. Flex is the
flexibility measure as constructed in Section 3 (i.e. Flext ¼ Max_flex t 2 CLDAt);
DGNP t & DCPI t measure the real economic growth and inflation, respectively; lpbit is
the firm’s market-to-book of equity ratio at the beginning of the quarter and proxies for its
expected future growth. I also include variables, other than the flexibility measure, that
354 likely affect a firm’s probability of missing analysts’ forecasts. These include:
(i) Pre_surp: the pre-managed earnings surprise or change. I use the current quarter
earnings excluding discretionary accruals as a proxy for pre-managed earnings.
Pre_surp is computed as pre-managed earnings of the current quarter less the
reported earnings of the same quarter in the prior year[13]. A firm’s probability of
successfully managing earnings to meet or beat analysts’ forecasts is a function
of both the flexibility available and the pre-managed earnings. A firm with a
positive pre-managed earnings change is more likely to meet the analysts’
forecast than a firm with a significant drop.
(ii) Exp_sur: the expected earnings change of the same quarter in the following year.
I use the consensus analysts’ forecast on earnings for the same quarter next year
as a proxy for expected earnings. Specifically, I use the first consensus analysts’
forecast released after current earnings announcement as a proxy. The expected
future earnings change is then computed as the difference between this
consensus forecast and the pre-managed earnings of the current quarter
(i.e. reported earnings stripped of discretionary accruals).
Managers consider expected future earnings when making earnings management

decisions (Fudenberg and Tirole, 1995; DeFond and Park, 1997; Elgers et al., 2003).
When the manager expects next period’s earnings to be lower than this period’s, he/she
has less incentive to “borrow” from the future. By “borrowing” from the future, the
manager simply defers the losses to the next period. This can increase the cost of
reporting the losses. For example, if reported losses in the subsequent period trigger
any shareholders’ lawsuit, the deferral of losses can increase the potential litigation
cost or settlement amount due to the extended class action period. I include the
expected future earnings change in the analysis to control for this difference in the
manager’s incentive to manage earnings.
Appendix provides a detailed description of variables used in the analyses.
I first perform the analysis using the sample period of Barton and Simko (2002) to
see if my flexibility measure has any impact on a firm’s probability of missing
analysts’ forecasts in their sample period. I then rerun the probit analysis using the full
sample. Although prior studies suggest that the effect of flexibility is likely to be the
greatest for firms with small misses, a sample selection based on the size of reported
forecast errors could omit those firms most constrained by flexibility. Firm-quarters
with low flexibility cannot even hide small negative pre-managed forecast errors.
Further, these firms have incentives to take a big bath in order to relax the flexibility
constraint for the following quarters. By excluding firms with large forecast errors,
I am likely to omit these firms. Consequently, I perform my analyses on all firms with
the required data. I have also repeated the analyses on a sub-sample of firm-quarters
with small absolute reported forecast errors and the results are qualitatively the same.
Hence, I do not report results of the sub-sample analyses.
Descriptive statistics Earnings
Table II Panel A provides descriptive statistics on firm-quarters that meet/beat management
analysts’ forecasts (meeting/beating firms) and those that miss the forecasts (missing
firms) in the full sample. A histogram of the distribution of reported forecasted errors flexibility
(not presented) shows that, similar to that documented in Burgstahler and Dichev
(1997) and Abarbanell and Lehavy (2003), there is a peak in the number of
firm-quarters with small positive forecast errors (i.e. firms with forecast errors in 355
(0, 0.01)). About 33 percent of the sample missed analysts’ forecasts.
On average, meeting/beating firms have an increase in pre-managed earnings
compared to the same quarter last year whereas missing firms have a decrease in
pre-managed earnings. Meeting/beating firms also tend to be larger (as measured by
total assets) than missing firms. Further, consistent with my hypothesis, the mean
flexibility available (Op_ flex) of the missing firms (0.130) is lower than that of the
meeting firms (0.135). The difference is significant at the 1 percent level. The difference
in flexibility can be attributed to the lower flexibility limits (meeting/beating firms:
0.132 vs missing firms: 0.128) and the higher level of flexibility used (meeting/beating
firms: 2 0.003 vs missing firms: 2 0.002) by the missing firms. Meeting firms have
a mean market-to-book ratio of 3.319 whereas that of missing firms is only 2.830.
The differences are significant at the 1 percent level. This lower probability of missing
analysts’ forecasts for high-growth firms can be due to their higher cost – more negative
returns to the negative surprise – of missing forecasts (Skinner and Sloan, 2002).
The change in GDP tends to be higher while the change in CPI tends to be lower in
the firm-quarters meeting/beating analysts’ forecasts than those missing the forecasts.
Panel B shows the correlation matrix among variables used in the test: Pearson
correlation is shown below the diagonal while Spearman is shown above the diagonal.
An examination of the Spearman correlation reveals that forecast error has a positive
correlation with the flexibility available (Op_ flex) and the flexibility limit (Max_ flex),
and a negative correlation with the flexibility used (CLDA). On the other hand, forecast
error is negatively correlated with the lagged change in total accruals (DTotal accrual )
and lagged net operating assets (lnoa). Forecast error is also positively correlated with
market-to-book ratio, pre-managed earnings surprise, and expected future earnings
change. Op_flex has a negative correlation with DTotal accrual but a positive
correlation with lnoa.
Regression results
Flexibility measures examined in separate regression analysis. Table III presents results
of the multivariate regression analysis using observations in the Barton and Simko
(2002) sample period (1993-1999; shown in Panel A) and those in the full sample period
(1990-2009; shown in Panel B). Since the coefficients in a probit regression do not reflect
the rate of change in the dependent variable as the independent variables change, the
marginal effect of each independent variable on the dependent variable is also
presented. The marginal effect of an independent variable can be interpreted the same
way as the OLS coefficients.
Panel A shows that when Barton and Simko’s measure (i.e. net operating assets
deflated by sales) is used, one unit increase in the lagged net operating assets increases
the probability of missing analysts’ forecasts by 0.003 (marginal effect) and is
significant at the 1 percent level (columns 1-2). On the other hand, Kasznik’s flexibility
11,4
356
RAF
Table II.
surprise test
correlation matrix for
Summary statistics and
variables in the earnings

(1) (2) (3) (4) (5) (6)
Firm-quarters meeting/beating Firm-quarters missing forecast
forecast (47,940 observations; (23,237 observations; 32.65 percent (8)
67.35 percent of sample) of sample) (7) Wilcoxon
Mean Median SD Mean Median SD t-test z-statistic
Panel A: comparison of the characteristics of firms that meet/beat analysts’ forecasts vs those of firms that miss the
forecasts
Exp_sur 20.391 2 0.381 0.456 20.391 2 0.382 0.438 0.1611 4.473 * * *
Pre_surp 0.002 0.001 0.048 20.006 2 0.005 0.052 20.968 * * * 24.131 * * *
Total assets 3,819.24 699.97 12,886.92 2,939.82 454.24 12,267.79 8.82 * * * 28.149 * * *
Lagged net
operating assets
(deflated by net
sales) 5.121 2.031 262.760 3.610 2.133 52.677 0.868 29.956 * * *
Operating cycle
(no. of quarters) 1.851 2.00 1.077 1.869 2.00 1.096 21.98 * * 21.538
Flexibility
available (Op_
flex) 0.135 0.115 0.138 0.130 0.110 0.131 4.732 * * * 9.808 * * *
Flexibility limit
(Max_flex) 0.132 0.113 0.126 0.128 0.110 0.117 3.821 * * * 6.155 * * *
Flexibility used
(CLDA) 20.003 2 0.001 0.053 20.002 0.001 0.055 23.200 * * * 27.968 * * *
Market-to-book
of equity ratio 3.319 2.416 3.690 2.830 2.056 3.346 17.114 * * * 27.956 * * *
DGDP (%) 2.193 2.673 2.004 2.137 2.514 1.959 3.509 * * * 5.000 * * *
DCPI (%) 2.557 2.787 1.149 2.713 2.848 1.177 216.824 * * * 218.263 * * *
(continued)
(1) (2) (3) (4) (5) (6)

Firm-quarters meeting/beating Firm-quarters missing forecast
forecast (47,940 observations; (23,237 observations; 32.65 percent (8)
67.35 percent of sample) of sample) (7) Wilcoxon
Mean Median SD Mean Median SD t-test z-statistic
Panel B: correlations among variables for the earnings surprise test

Forecast Op_flex max_flex CLDA DTotal lnoa lpb DGDP DCPI Pre_ Exp_sur
error accrual surp
Forecast error 1 0.026 0.008 20.034 2 0.007 20.044 0.04 20.039 20.049 0.105 0.011
(0.000) (0.026) (0.000) (0.075) (0.000) (0.000) (0.000) (0.000) (0.000) (0.003)
Op_flex 0.001 1 0.832 20.049 2 0.016 0.012 0.042 20.102 20.07 20.02 2 0.001
(0.775) (0.000) (0.000) (0.000) (0.002) (0.000) (0.000) (0.000) (0.000) (0.713)
max_flex 0.001 0.918 1 20.015 2 0.026 0.033 0.058 20.097 20.081 20.02 2 0.003
(0.878) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.468)
CLDA 20.001 2 0.422 2 0.029 1 2 0.01 0.048 0.005 0.026 0.000 0.002 0.014
(0.712) (0.000) (0.000) (0.007) (0.007) (0.181) (0.000) (0.915) (0.516) (0.000)
DTotal accrual 0 2 0.002 0 0.004 1 0.045 20.009 20.024 0.001 0.054 2 0.169
(0.900) (0.589) (0.905) (0.274) (0.000) (0.022) (0.000) (0.713) (0.000) (0.000)
lnoa 0.001 0.003 0.004 0.001 0 1 20.143 20.015 20.013 0.010 2 0.020
(0.884) (0.403) (0.290) (0.758) (0.993) (0.000) (0.000) (0.001) (0.009) (0.000)
lpb 0.015 0.032 0.039 0.007 2 0.003 0 1 0.151 0.047 0.020 0.036
(0.000) (0.000) (0.000) (0.055) (0.416) (0.917) (0.000) (0.000) (0.000) (0.000)
DGDP 0.001 2 0.094 2 0.083 0.047 0 20.003 0.101 1 20.043 0.041 0.029
(0.713) (0.000) (0.000) (0.000) (0.902) (0.354) (0.000) (0.000) (0.000) (0.000)
DCPI 20.014 2 0.055 2 0.044 0.037 0.009 0.004 0.042 0.403 1 20.023 0.000
(0.000) (0.000) (0.000) (0.000) (0.015) (0.338) (0.000) (0.000) (0.000) (0.966)
Pre_surp 0.026 2 0.012 2 0.006 0.017 0.003 20.002 0.013 0.033 20.009 1 2 0.433
(0.000) (0.002) (0.139) (0.000) (0.467) (0.568) (0.000) (0.000) (0.016) (0.000)
Exp_sur 20.29 0.004 0.004 0 0 0 20.003 0.015 20.009 20.051 1
(0.000) (0.326) (0.266) (0.947) (0.945) (0.937) (0.388) (0.000) (0.019) (0.000)
(continued)
management
Earnings
flexibility
357
Table II.
11,4
358
RAF
Table II.
Notes: Panel A: significant at: *10, * *5 and * * *1 percent levels (two-tailed); this table presents summary statistics of the 71,177 sample firm-quarters used
in the analysts’ forecast analyses; only firm-quarters with the consensus analysts’ forecast for the sample quarter and for the same quarter next year (required
for the computation of expected earnings change) are included in the sample; Panel A presents the descriptive statistics while Panel B presents the correlation
matrix of the variables; Panel A columns 1-3 present the statistics for firm-quarters that meet/beat analysts’ forecasts while columns 4-6 present those for
firm-quarters that miss the analysts’ forecasts; column 7 presents the t-statistics for the test on the difference between the firm-quarters that meet/beat
analysts’ consensus forecasts and those that miss the forecasts; column 8 presents the Wilcoxon z-statistics for the difference; Panel B: this table presents the
Pearson (below the diagonal) and Spearman (above the diagonal) correlation among the variables; p-values are presented in parenthesis below the correlation
coefficients; where, Forecast error – difference between actual earnings per share and the last consensus analysts’ forecast before earnings announcement;
Exp_sur – expected future earnings change; this expected earnings change is measured by subtracting earnings stripped of discretionary accruals of the
current quarter from the expected earnings of the same quarter of the following year, t þ 4 [i.e. (E[TEtþ 4] 2 REt þ DA_adjustedt)]; expected earnings are
proxied by the first consensus analysts’ forecast of t þ 4 released immediately after the earnings announcement of t; Pre_surp – the pre-managed earnings
change of the current quarter deflated by lagged total assets; the pre-managed earnings of a quarter are defined as earnings before extraordinary items and
discontinued operations less discretionary accruals estimated by the cross-sectional Jones model adjusted for sales growth and dividend payout ratio; the pre-
managed earnings change is then calculated as the difference between the current pre-managed earnings and the reported earnings of the same quarter
last year [i.e. (REt 2 DAt 2 REt2 4)]; Op_flex – the cross-sectional operating cycle flexibility available, which is calculated as the upper flexibility bound
(Max_flex) less the sum of discretionary accruals cumulated over Qi quarters (CLDA), where Qi is the number of quarters in the firm’s operating cycle; a
firm’s operating cycle is proxied by the mean industry operating cycle, computed using the operating cycles of all firms in the same quarter and industry; the
discretionary accruals used to construct Op_flex are those estimated by:
TAt a DSalest PPE t
¼ þ b1 þ b2 þ b3 SGt þ b4 Divid t þ 1t ð2Þ
TAsset t21 TAsset t21 TAsset t21 TAssett21
The upper flexibility bound, which is calculated as the mean of the industry’s (as defined by two-digit SIC) cumulative lagged discretionary accruals plus two times the
standard deviations of these cumulative lagged accruals; Max_flex – the upper flexibility bound, which is calculated as the mean of the industry’s (as defined by two-
digit SIC) cumulative lagged discretionary accruals plus two times the standard deviations of these cumulative lagged accruals; CLDA – flexibility used to manage
earnings in the prior quarters and have not yet reversed at the beginning of the current quarter; it is calculated as the sum of discretionary accruals cumulated over Qi
quarters, where Qi is the number of quarters in the firm’s operating cycle; DTotal accrual – the difference between the total accruals of this quarter and those of the
same quarter last year, deflated by total accruals of the same quarter last year; total accruals are computed as net income before extraordinary items less net operating
cash flows; lnoa – lagged net operating assets deflated by sales; net operating assets are defined as shareholders’ equity less cash and marketable securities, plus debt;
lpb – lagged market value of equity deflated by book value of equity; DGDP – percentage change in real GDP; DCPI – percentage change in consumer price index
(1) (2) (3) (4) (5) (6)

Barton and Simko’s measure Kasznik’s measure CS operating cycle
Predicted sign Coefficient (SE) Marginal effects Coefficient (SE) Marginal effects Coefficient (SE) Marginal effects
Panel A: Barton and Simko’s sample period (1993-1999)

lnoa þ 0.008 * * * 0.003
(0.003)
DTotal accrual þ 0.000 0.000
(0.000)
Op_flex 2 2 0.307 * * 2 0.113
(0.144)
DGDP 2 2 0.117 * * * 2 0.043 2 0.117 * * * 2 0.043 2 0.117 * * * 2 0.043
(0.017) (0.017) (0.017)
DCPI þ 0.059 * * 0.022 0.059 * * 0.022 0.057 * * 0.021
(0.028) (0.028) (0.028)
lpb 2 2 0.026 * * * 2 0.009 2 0.026 * * * 2 0.010 2 0.026 * * * 2 0.009
(0.003) (0.003) (0.003)
Pre_surp 2 2 2.492 * * * 2 0.920 2 2.500 * * * 2 0.922 2 2.510 * * * 2 0.926
(0.213) (0.213) (0.213)
Exp_sur 2 2 0.475 * * * 2 0.175 2 0.486 * * * 2 0.179 2 0.481 * * * 2 0.177
(0.122) (0.122) (0.122)
Intercept 2 0.195 2 0.177 2 0.137
(0.135) (0.135) (0.136)
Observations 17,315 17,315 17,315
McFadden R 2 0.0168 0.0164 0.0165
Panel B: full sample period (1990-2009)
lnoa þ 2 0.000 2 0.000
(0.000)
(0.000)
Op_flex 2 2 0.159 * * * 2 0.057
(0.038)
(continued)
management
Earnings
analysts’ forecasts
probability of missing
flexibility
Probit analysis of the

Table III.
359
11,4
360
RAF
Table III.
(1) (2) (3) (4) (5) (6)
DGDP 2 2 0.023 * * * 2 0.008 2 0.023 * * * 2 0.008 2 0.024 * * * 2 0.009

(0.003) (0.003) (0.003)
DCPI þ 0.089 * * * 0.032 0.088 * * * 0.032 0.088 * * * 0.032
(0.005) (0.005) (0.005)
lpb 2 2 0.024 * * * 2 0.009 2 0.024 * * * 2 0.009 2 0.024 * * * 2 0.009
(0.001) (0.001) (0.001)
Pre_surp 2 2 2.012 * * * 2 0.724 2 2.012 * * * 2 0.724 2 2.017 * * * 2 0.725
(0.100) (0.100) (0.100)
Exp_sur 2 2 0.010 2 0.003 2 0.010 2 0.003 2 0.009 2 0.003
(0.010) (0.010) (0.010)
Intercept 2 0.565 * * * 2 0.565 * * * 2 0.541 * * *
(0.013) (0.013) (0.015)
Observations 71,177 71,177 71,177
McFadden R 2 0.0123 0.0123 0.0125
Notes: Panel A: significanct at: *10 * *5 and * * *1 percent levels (two-tailed); probit regressions investigating relation between a firm’s probability of missing analysts’ forecasts and its
earnings management flexibility:
Prðmissit ¼ 1Þ ¼ Fða þ b1 Flexit þ b2 DGNP t þ b3 DCPI t þ b4 lpbit þ b5 Pre_surpit þ b6 Exp_surit Þ:
The marginal effect of x (independent variable) on y (dependent variable) is also provided; Panel A presents the results for the sample period of Barton and Simko (2002) – 1993-1999; Panel B
presents the results for my sample period (1990-2009); Panel B: where: Miss – a dummy variable that takes on a value of 1 if the firm’s reported earnings are below the last consensus
analysts’ forecast before the earnings announcement, 0 otherwise; lnoa – lagged net operating assets deflated by sales; Net operating assets are defined as shareholders’ equity less cash and
marketable securities, plus debt; DTotal accrual – the difference between the total accruals of this quarter and those of the same quarter last year, deflated by total accruals of the same
quarter last year; total accruals are computed as net income before extraordinary items less net operating cash flows; Op_flex – the cross-sectional operating cycle flexibility available, which
is calculated as the upper flexibility bound (Max_flex) less the sum of discretionary accruals cumulated over Qi quarters (CLDA), where Qi is the number of quarters in the firm’s operating
cycle; a firm’s operating cycle is proxied by the mean industry operating cycle, computed using the operating cycles of all firms in the same quarter and industry; the discretionary accruals
used to construct Op_flex are those estimated by:
TAt a DSalest PPE t
TAssett21 TAssett21 TAsset t21 TAsset t21
The upper flexibility bound, which is calculated as the mean of the industry’s (as defined by two-digit SIC) cumulative lagged discretionary accruals plus two times the standard deviations of
these cumulative lagged accruals; DGDP – percentage change in real GDP; DCPI – percentage change in consumer price index; lpb – lagged market value of equity deflated by book value of
equity; Pre_surp – the pre-managed earnings change of the current quarter deflated by lagged total assets; the pre-managed earnings of a quarter are defined as earnings before
extraordinary items and discontinued operations less discretionary accruals estimated by the cross-sectional Jones model adjusted for sales growth and dividend payout ratio; the pre-
managed earnings change is then calculated as the difference between the current pre-managed earnings and the reported earnings of the same quarter last year [i.e. (REt 2 DAt 2 REt2 4)];
Exp_sur – expected future earnings change; this expected earnings change is measured by subtracting earnings stripped of discretionary accruals of the current quarter from the expected
earnings of the same quarter of the following year, t þ 4 [i.e. (E [TEtþ 4] 2 REt þ DA_adjustedt)]; expected earnings are proxied by the first consensus analysts’ forecast of t þ 4 released
immediately after the earnings announcement of t
proxy (i.e. lagged change in total accruals) has no significant effect (columns 3-4). Earnings
When the cross-sectional operating cycle flexibility measure is used, it has a marginal management
effect of 2 0.113 (significant at the 1 percent level) on the probability of missing
analysts’ forecasts (columns 5-6). These results suggest that the marginal effect of the flexibility
cross-sectional operating cycle measure is more than 30 times that of Barton and
Simko’s measure.
Most of the coefficients of the control variables have the expected sign. The change in 361
GDP (DGDP), the lagged market-to-book ratio (lpb), the pre-managed earnings change
(Pre_surp), and the expected future earnings change all have significant negative
coefficients in the probit analysis[14].The coefficient of the inflation indicator (DCPI ) is
significantly positive. Results based on a sub-sample of firms with forecast errors less
than 1 percent (in absolute term) of their respective stock price at the beginning of the
quarter are qualitatively the same as those of the full sample and are not reported[15].
Panel B presents the results when observations in the full sample period (i.e. 1990-2009)
are used in the analyses. The main difference between this set of results and those in Panel
A is that the correlation between Barton and Simko’s measure and the probability of
missing analysts’ forecast is no longer significant in this extended sample period.
Flexibility measures examined in the same regression analysis

To examine whether my measure provides incremental information to that of Barton and
Simko, I include both measures simultaneously in the analysis. As both measures are
closely tied to a firm’s performance in prior periods, the effect of flexibility on the
probability of missing analysts’ forecasts reported above can simply be driven by the
firm’s past performance. Hence, I control for a firm’s ROA in the prior eight quarters in
the following analysis[16].
Results in Table IV shows that, similar to those in Table III Panel B, except for my
measure (i.e. the cross-sectional operating cycle flexibility measure), the other flexibility
measures do not have any significant correlation with a firm’s probability of missing
analysts’ forecast. When I include both my measure and Barton and Simko’s measure
(columns 7-8) or Kasznik’s measure (columns 9-10) simultaneously in the model, only my
measure has a significant negative correlation with the probability. Coefficients of all
other variables are similar to those reported in Table III Panel B. The first four lags of
ROA are all negatively correlated with a firm’s probability of missing analysts’ forecast,
suggesting that firms with good performance in the preceding year are less likely to miss
analysts’ forecasts.
Industry-adjusted flexibility measures

As DeFond (2002) suggest that Barton and Simko’s measure can be capturing the
industry effect in asset turnover ratio rather than a firm’s flexibility used in prior
periods, I repeat the analyses using flexibility measures adjusted by the industry mean
(i.e. a firm’s available earnings management flexibility – mean of earnings management
flexibility of the industry). I use the two-digit SIC code to define the industry. I do not
simply add industry dummies in the probit analysis because coefficients of these
variables capture only the intercept effect, but not the slope effect, of the different
industries. Since I want to examine whether the firm-specific component of the flexibility
measure has an impact on the probability of missing analysts’ forecasts (i.e. the slope
effect), I use the industry-adjusted flexibility measures in the analysis.
11,4
362
RAF
Table IV.
performance
Analysis of the
analysts’ forecasts –
controlling for lagged
(1) (2) (7) (8) (9) (10)
Barton and Simko’s (3) (4) (5) (6) Both Barton and Simko and Both Kasznik and CS operating
measure Kasznik’s measure CS operating cycle CS operating cycle cycle
Predicted Coefficient Marginal Coefficient Marginal Coefficient Marginal Coefficient Marginal Coefficient Marginal
sign (SE) effects (SE) effects (SE) effects (SE) effects (SE) effects
lnoa þ 2 0.000 2 0.000 2 0.000 2 0.000

(0.000) (0.000)
DTotal þ 0.000 0.000 0.000 0.000
accrual
(0.000) (0.000)
Op_flex 2 2 0.291 * * * 2 0.104 2 0.291 * * * 2 0.104 2 0.291 * * * 2 0.104
(0.040) (0.040) (0.040)
DGDP 2 2 0.019 * * * 2 0.007 2 0.019 * * * 2 0.007 2 0.020 * * * 2 0.007 2 0.020 * * * 2 0.007 2 0.020 * * * 2 0.007
(0.003) (0.003) (0.003) (0.003) (0.003)
DCPI þ 0.095 * * * 0.034 0.095 * * * 0.034 0.095 * * * 0.034 0.095 * * * 0.034 0.095 * * * 0.034
(0.005) (0.005) (0.005) (0.005) (0.005)
lpb 2 2 0.023 * * * 2 0.008 2 0.023 * * * 2 0.008 2 0.023 * * * 2 0.008 2 0.023 * * * 2 0.008 2 0.023 * * * 2 0.008
(0.001) (0.001) (0.001) (0.001) (0.001)
Pre_surp 2 2 2.050 * * * 2 0.736 2 2.049 * * * 2 0.736 2 2.056 * * * 2 0.738 2 2.057 * * * 2 0.738 2 2.057 * * * 2 0.738
(0.103) (0.103) (0.103) (0.103) (0.103)
Exp_sur 2 2 0.008 2 0.003 2 0.008 2 0.003 2 0.007 2 0.003 2 0.007 2 0.003 2 0.007 2 0.003
(0.010) (0.010) (0.010) (0.010) (0.010)
LROA 2 2 1.358 * * * 2 0.488 2 1.356 * * * 2 0.487 2 1.419 * * * 2 0.509 2 1.421 * * * 2 0.510 2 1.419 * * * 2 0.509
(0.125) (0.125) (0.125) (0.126) (0.125)
L2ROA 2 2 0.593 * * * 2 0.213 2 0.592 * * * 2 0.212 2 0.689 * * * 2 0.247 2 0.689 * * * 2 0.247 2 0.689 * * * 2 0.247
(0.129) (0.129) (0.130) (0.130) (0.130)
L3ROA 2 2 0.255 * 2 0.092 2 0.254 * 2 0.091 2 0.231 * 2 0.083 2 0.232 * 2 0.083 2 0.231 * 2 0.083
(0.132) (0.132) (0.133) (0.133) (0.133)
L4ROA 2 2 0.490 * * * 2 0.176 2 0.487 * * * 2 0.175 2 0.501 * * * 2 0.180 2 0.504 * * * 2 0.181 2 0.501 * * * 2 0.180
(0.132) (0.132) (0.132) (0.132) (0.132)
L5ROA 2 0.155 0.056 0.155 0.056 0.165 0.059 0.165 0.059 0.165 0.059
(0.106) (0.106) (0.106) (0.106) (0.106)
L6ROA 2 0.040 0.014 0.041 0.015 0.051 0.018 0.049 0.018 0.051 0.018
(0.094) (0.094) (0.094) (0.094) (0.094)
L7ROA 2 0.066 0.024 0.066 0.024 0.068 0.025 0.068 0.024 0.068 0.024
(0.090) (0.090) (0.091) (0.091) (0.091)
(continued)
(1) (2) (7) (8) (9) (10)

Barton and Simko’s (3) (4) (5) (6) Both Barton and Simko and Both Kasznik and CS operating
measure Kasznik’s measure CS operating cycle CS operating cycle cycle
Predicted Coefficient Marginal Coefficient Marginal Coefficient Marginal Coefficient Marginal Coefficient Marginal
sign (SE) effects (SE) effects (SE) effects (SE) effects (SE) effects
L8ROA 2 0.056 0.020 0.052 0.019 0.048 0.017 0.053 0.019 0.048 0.017
(0.088) (0.088) (0.088) (0.089) (0.088)
Intercept 2 0.579 * * * 2 0.579 * * * 2 0.537 * * * 2 0.537 * * * 2 0.537 * * *
(0.014) (0.014) (0.015) (0.015) (0.015)
Observations 70,405 70,405 70,405 70,405 70,405
McFadden 0.0176 0.0176 0.0182 0.0182 0.0182
R2
Notes: Significanct at: *10, * *5 and * * *1 percent levels (two-tailed); this table examines the relation between the probability of missing analysts’ forecasts and earnings management
flexibility, after controlling for lagged performance; where: Miss – a dummy variable that takes on a value of 1 if the firm’s reported earnings are below the last consensus analysts’
forecast before the earnings announcement, 0 otherwise; lnoa – lagged net operating assets deflated by sales; Net operating assets are defined as shareholders’ equity less cash and
marketable securities, plus debt; DTotal accrual – the difference between the total accruals of this quarter and those of the same quarter last year, deflated by total accruals of the same
quarter last year; Total accruals is computed as net income before extraordinary items less net operating cash flows; Op_flex – the cross-sectional operating cycle flexibility available,
which is calculated as the upper flexibility bound (Max_flex) less the sum of discretionary accruals cumulated over Qi quarters (CLDA), where Qi is the number of quarters in the firm’s
operating cycle; a firm’s operating cycle is proxied by the mean industry operating cycle, computed using the operating cycles of all firms in the same quarter and industry; the
discretionary accruals used to construct Op_flex are those estimated by:
TAt a DSalest PPE t
The upper flexibility bound, which is calculated as the mean of the industry’s (as defined by two-digit SIC) cumulative lagged discretionary accruals plus two times the standard
deviations of these cumulative lagged accruals; DGDP – percentage change in real GDP; DCPI – percentage change in consumer price index; lpb – lagged market value of equity deflated
by book value of equity; Pre_surp – the pre-managed earnings change of the current quarter deflated by lagged total assets; the pre-managed earnings of a quarter are defined as earnings
before extraordinary items and discontinued operations less discretionary accruals estimated by the cross-sectional Jones model adjusted for sales growth and dividend payout ratio; the
pre-managed earnings change is then calculated as the difference between the current pre-managed earnings and the reported earnings of the same quarter last year [i.e.
(REt 2 DAt 2 REt2 4)]; Exp_sur – expected future earnings change; this expected earnings change is measured by subtracting earnings stripped of discretionary accruals of the current
quarter from the expected earnings of the same quarter of the following year, t þ 4 [i.e. (E [TEtþ 4] 2 REt þ DA_adjustedt)]; expected earnings are proxied by the first consensus analysts’
forecast of t þ 4 released immediately after the earnings announcement of t; LROA – ROA of last quarter (i.e. t 2 1); ROA is calculated as income before extraordinary items deflated by
assets at the beginning of the quarter; L2ROA,. . .,L8ROA – ROA of quarter t 2 2,. . .t 2 8
management
Earnings
flexibility
363
Table IV.
RAF Table V Panel A presents the results using these industry-adjusted measures while Panel
11,4 B presents results of the same analyses after controlling for lagged ROAs. Columns 1-2 of
Panel A show the results when the observations are restricted to those in the Barton and
Simko sample period while the remaining columns present those when the full sample
period is used[17]. Results suggest that after adjusting for the industry mean, Barton and
Simko’s measure no longer has any significant correlation with a firm’s probability of
364 missing analysts forecast in either of the sample periods. Columns 5-6 indicate that
industry-adjusted Kasznik’s measure also does not have any significant impact. In fact,
the only measure that has a significant correlation with a firm’s probability of missing
analysts’ forecast is my measure (columns 7-8), which has a marginal effect of 20.040.
These results confirm the suspicion of DeFond (2002) and suggest that Barton and Simko’s
measure mainly reflects the variation in asset turnover ratios (their flexibility measure)
across industries, rather than a firm’s earnings management flexibility.
Panel B presents the results after controlling for the effect of lagged performance.
Results are similar to those in Panel A. When I include both my measure and Barton and
Simko’s measure (columns 7-8) or Kasznik’s measure (columns 9-10), only my measure
has a significant impact.
Overall, results suggest my flexibility measure has a stronger correlation with
a firm’s probability of missing analysts forecast than these other measures proposed
in prior studies. Also, my measure captures certain aspects of a firm’s
earnings management flexibility – variation over time and across firms within the
same industry – that are not reflected in the other measures.
7. Additional sensitivity analysis

The Sarbanes-Oxley Act (SOX) was introduced during the sample period. The objective of
the SOX is to restore integrity in the financial market by reinforcing corporate
accountability and improving the accuracy and reliability of corporate disclosures.
The SOX is likely to have an impact on firms’ earnings management flexibility. Cohen et al.
(2008) and Bartov and Cohen (2008) both document a decline in accrual management in the
post-SOX period. This decline in accrual management can be caused by a decline in
the earnings management flexibility (Choy, 2010). To examine whether this change in the
regulatory environment has any effect on the relationship between earnings management
flexibility and a firm’s probability of missing analysts’ forecast, I add interaction terms
between SOX and all the variables into my model. These interaction terms capture the
change in the impact of these variables on a firm’s probability of missing analysts’
forecasts in the post-SOX period.
Results in Table VI indicate that none of the interaction terms between SOX and the
flexibility measure – Barton and Simko’s, Kasznik’s and mine – has a statistically
significant coefficient, suggesting that there is no significant change in the correlation
between flexibility and the probability of missing analysts’ forecast after the
implementation of the SOX.
In addition to the change in regulatory environment, a firm’s auditor also plays an
important monitoring role in constraining earnings management. Prior studies have
employed the BigN as a proxy for auditor reputation and found firms willing to pay a
premium for reputation (Francis, 1984; Francis and Stokes, 1986; Francis and Simon, 1987;
Palmrose, 1986). I include a dummy variable, BigN, to control for any effect this perceived
premium audit quality can have on the probability of missing analysts’ forecasts.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Barton and Simko’s measure

and sample period Barton and Simko’s measure Kasznik’s measure CS operating cycle
Predicted Marginal Marginal Marginal Marginal
sign Coefficient (SE) effects Coefficient (SE) effects Coefficient (SE) effects Coefficient (SE) effects
Panel A: impact of industry-adjusted flexibility and firms’ probability of missing analysts’ forecast
Ind_lnoa þ 0.001 0.001 20.000 2 0.000
(0.001) (0.000)
Ind_DTotal þ 0.000 0.000
accrual
(0.000)
Ind_Op_flex 2 2 0.111 * * * 2 0.040
(0.039)
DGDP 2 2 0.117 * * * 2 0.043 20.023 * * * 2 0.008 2 0.023 * * * 2 0.008 2 0.023 * * * 2 0.008
(0.017) (0.003) (0.003) (0.003)
DCPI þ 0.059 * * 0.022 0.089 * * * 0.032 0.088 * * * 0.032 0.088 * * * 0.032
(0.028) (0.005) (0.005) (0.005)
Lpb 2 2 0.026 * * * 2 0.009 20.024 * * * 2 0.009 2 0.024 * * * 2 0.009 2 0.024 * * * 2 0.009
(0.003) (0.001) (0.001) (0.001)
Pre_surp 2 2 2.499 * * * 2 0.922 22.012 * * * 2 0.724 2 2.012 * * * 2 0.724 2 2.015 * * * 2 0.725
(0.213) (0.100) (0.100) (0.100)
Exp_sur 2 2 0.488 * * * 2 0.180 20.010 2 0.003 2 0.010 2 0.003 2 0.010 2 0.003
(0.122) (0.010) (0.010) (0.010)
Intercept 2 0.175 20.565 * * * 2 0.565 * * * 2 0.563 * * *
(0.135) (0.013) (0.013) (0.013)
Observations 17,315 71,177 71,177 71,177
2
McFadden R 0.0164 0.0123 0.0123 0.0124
Both Barton and Simko and Both Kasznik and CS
Barton and Simko’s measure Kasznik’s measure CS operating cycle CS operating cycle operating cycle
Predicted Marginal Marginal Marginal Marginal Marginal
sign Coefficient (SE) effects Coefficient (SE) effects Coefficient (SE) effects Coefficient (SE) effects Coefficient (SE) effects
Panel B: impact of industry-adjusted flexibility and firms’ probability of missing analysts’ forecast – control for lagged performance
Ind_lnoa þ 2 0.000 2 0.000 2 0.000 2 0.000
(0.000) (0.000)
Ind_DTotal þ 0.000 0.000 0.000 0.000
accrual (0.000) (0.000)
Ind_Op_flex 2 2 0.213 * * * 2 0.076 2 0.213 * * * 2 0.076 2 0.212 * * * 2 0.076
(0.041) (0.041) (0.041)
DGDP 2 2 0.019 * * * 2 0.007 20.019 * * * 2 0.007 2 0.019 * * * 2 0.007 2 0.019 * * * 2 0.007 2 0.019 * * * 2 0.007
(0.003) (0.003) (0.003) (0.003) (0.003)
(continued)
using industry-adjusted
management
Earnings
flexibility
flexibility measures
Analysis of the
Table V.
365
11,4
366
RAF
Table V.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
DCPI þ 0.095 * * * 0.034 0.095 * * * 0.034 0.095 * * * 0.034 0.095 * * * 0.034 0.095 * * * 0.034
(0.005) (0.005) (0.005) (0.005) (0.005)
Lpb 2 2 0.023 * * * 2 0.008 20.023 * * * 2 0.008 2 0.023 * * * 2 0.008 2 0.023 * * * 2 0.008 2 0.023 * * * 2 0.008
(0.001) (0.001) (0.001) (0.001) (0.001)
Pre_surp 2 2 2.050 * * * 2 0.736 22.049 * * * 2 0.736 2 2.051 * * * 2 0.736 2 2.052 * * * 2 0.736 2 2.052 * * * 2 0.736
(0.103) (0.103) (0.103) (0.103) (0.103)
Exp_sur 2 2 0.008 2 0.003 20.008 2 0.003 2 0.007 2 0.003 2 0.007 2 0.003 2 0.007 2 0.003
(0.010) (0.010) (0.010) (0.010) (0.010)
LROA 2 2 1.358 * * * 2 0.487 21.356 * * * 2 0.487 2 1.395 * * * 2 0.501 2 1.397 * * * 2 0.501 2 1.395 * * * 2 0.501
(0.125) (0.125) (0.125) (0.125) (0.125)
L2ROA 2 2 0.592 * * * 2 0.213 20.592 * * * 2 0.213 2 0.658 * * * 2 0.236 2 0.658 * * * 2 0.236 2 0.658 * * * 2 0.236
(0.129) (0.129) (0.130) (0.130) (0.130)
L3ROA 2 2 0.255 * 2 0.092 20.254 * 2 0.091 2 0.233 * 2 0.084 2 0.234 * 2 0.084 2 0.233 * 2 0.084
(0.132) (0.132) (0.132) (0.132) (0.132)
L4ROA 2 2 0.490 * * * 2 0.176 20.488 * * * 2 0.175 2 0.492 * * * 2 0.177 2 0.495 * * * 2 0.178 2 0.492 * * * 2 0.177
(0.132) (0.132) (0.132) (0.132) (0.132)
L5ROA 2 0.155 0.056 0.155 0.056 0.165 0.059 0.165 0.059 0.165 0.059
(0.106) (0.106) (0.106) (0.106) (0.106)
L6ROA 2 0.040 0.014 0.041 0.015 0.051 0.018 0.050 0.018 0.051 0.018
(0.094) (0.094) (0.094) (0.094) (0.094)
L7ROA 2 0.066 0.024 0.066 0.024 0.070 0.025 0.070 0.025 0.070 0.025
(0.090) (0.090) (0.091) (0.091) (0.091)
L8ROA 2 0.057 0.020 0.052 0.019 0.053 0.019 0.058 0.021 0.053 0.019
(0.088) (0.088) (0.088) (0.089) (0.088)
Intercept 2 0.580 * * * 20.579 * * * 2 0.577 * * * 2 0.577 * * * 2 0.577 * * *
(0.014) (0.014) (0.014) (0.014) (0.014)
Observations 70,405 70,405 70,405 70,405 70,405
McFadden R 2 0.0176 0.0176 0.0179 0.0179 0.0179
Notes: Significant at: *10, * *5 and * * *1 percent levels (two-tailed); this table examines the relation between industry-adjusted flexibility measures and the
probability of missing analysts’ forecasts; instead of the raw measure of flexibility, the deviation of each firm’s flexibility measure from its industry mean is used in
the probit analysis; Panel A presents the results without controlling for the lagged performance whereas Panel B presents those after controlling for lagged
performance; columns 1-2 of Panel A present results for the sample period of Barton and Simko (2002), 1993-1999, while columns 3-8 present those for the full sample
period (1990-2009). where: Ind_lnoa, Ind_DTotal accrual, and Ind_Op_flex are computed as the deviation of the respective flexibility measure (i.e. lnoa, DTotal
accrual, and Op_flex,) from their industry mean in each quarter; industry is defined by two-digit SIC code; LROA – ROA of last quarter (i.e. t 2 1). ROA is
calculated as income before extraordinary items deflated by assets at the beginning of the quarter; L2ROA,. . ., L8ROA – ROA of quarter t 2 2,. . .t 2 8; all other
variables are as defined in Table III
(1) (2) (3) (4) (5) (6)

lnoa þ 0.000 0.000

(0.000)
(0.000)
Op_flex 2 20.170 * * * 20.061
(0.044)
DGDP 2 20.022 * * * 2 0.008 2 0.022 * * * 20.008 20.024 * * * 20.009
(0.004) (0.004) (0.005)
DCPI þ 0.141 * * * 0.051 0.141 * * * 0.051 0.141 * * * 0.051
(0.007) (0.007) (0.007)
lpb 2 20.030 * * * 2 0.011 2 0.030 * * * 20.011 20.029 * * * 20.011
(0.002) (0.002) (0.002)
Pre_surp 2 21.973 * * * 2 0.709 2 1.975 * * * 20.710 21.983 * * * 20.713
(0.153) (0.153) (0.153)
Exp_sur 2 20.030 2 0.011 2 0.032 20.011 20.041 20.015
(0.051) (0.051) (0.052)
SOX*lnoa ? 20.000 2 0.000
(0.000)
SOX*DTotal accrual ? 2 0.000 20.000
(0.000)
SOX*Op_flex ? 0.027 0.010
(0.079)
SOX*DGDP ? 20.023 * * * 2 0.008 2 0.023 * * * 20.008 20.021 * * * 20.008
(0.007) (0.007) (0.007)
SOX*DCPI ? 20.059 * * * 2 0.021 2 0.059 * * * 20.021 20.061 * * * 20.022
(0.008) (0.008) (0.008)
SOX*lpb ? 0.012 * * * 0.004 0.012 * * * 0.004 0.011 * * * 0.004
(0.003) (0.003) (0.003)
SOX*Pre_surp ? 20.052 2 0.019 2 0.049 20.018 20.047 20.017
(0.203) (0.203) (0.203)
SOX*Exp_sur ? 0.020 0.007 0.021 0.008 0.031 0.011
(0.052) (0.052) (0.052)
Intercept 20.612 * * * 2 0.612 * * * 20.588 * * *
(0.015) 71,177 (0.015) 71,177 (0.018) 71,177
0.0169 0.0168 0.0170
Observations 71,177 71,177 71,177
McFadden R 2 0.0169 0.0168 0.0170
Notes: Significant at: *10, * *5 and * * *1 percent levels (two-tailed); this table compares the relation between the probability of missing analysts’ forecasts and earnings management flexibility in the
pre- and post-SOX periods
management
Earnings
pre- vs post-SOX
flexibility

367
Table VI.
RAF Results in Table VII indicate that after controlling for audit quality, the cross-sectional
11,4 operating cycle flexibility continues to have a significantly negative coefficient while
coefficients of the other two measures are insignificant. These results suggest that
findings in this study are robust to the inclusion of controls for audit quality and the
change in legal environment.
A firm’s probability of missing analysts’ forecast depends on whether it has enough
368 flexibility to cover the negative pre-managed earnings surprise. I construct a dummy
variable that captures whether a RE t firm can meet the analysts’ forecasts based on its
flexibility available and pre-managed earnings surprise (i.e. Reported Earnings (REt) –
Discretionary Accruals (DAt) – Forecast (Forecastt)). The dummy variable, Pmeet,
takes a value of 1 if the flexibility is large enough to cover the pre-managed negative
surprises[18] [i.e.Flext þ ðRE t 2 DAt 2 Forecast t Þ $ 0], zero otherwise. I rerun the
probit analysis with Pmeet replacing Op_flex and Pre_surp. Untabulated results show
that Pmeet has a significantly negative coefficient.
While the above results show that operating cycle flexibility measure does affect a
firm’s probability of missing analysts’ forecasts, it is not clear whether both the flexibility
limit and the flexibility used in prior periods have a significant impact. To address this
issue, I analyze the effect of flexibility limit and flexibility used separately. Untabulated
results show that the flexibility limit has a significantly negative effect (marginal effect
of 20.046) while the flexibility used has a significantly positive effect (marginal effect
of 0.110). These results suggest that both the flexibility limit and the flexibility used are
significant determinants of a firm’s ability to meet analysts’ forecast.
I also rerun the above analysis using different proxies for expected growth (e.g. actual
growth of the firm in the following year, sales growth in the prior five years, and
analysts’ forecast of long-term growth, etc.) and expected earnings of the current quarter
(e.g. earnings of the same quarter last year). Results are similar to those reported.
8. Conclusion
This study proposes a new measure of earnings management flexibility and compares it
with the flexibility measures proposed in Barton and Simko (2002) and Kasznik (1999).
Flexibility measures proposed in these two studies focus on the effect of prior earnings
management, whereas the measure constructed in this study accounts for the impact of
both the flexibility limits and the flexibility used. Specifically, this paper shows that the
limits of the allowable set of accruals play a critical role, in addition to prior earnings
management practices, in determining a firm’s available earnings management flexibility.
Results show that both Barton and Simko’s measure and my cross-sectional
operating cycle flexibility measure have a significant impact on a firm’s ability to
meet/beat analysts’ forecasts. When both measures are included in the same analysis,
the effect of my measure dominates. Further analysis shows that Barton and Simko’s
measure captures mainly the industry component of earnings management flexibility
(or the difference in operating characteristics across industries) while my measure
reflects both the industry and the firm-specific components of earnings management
flexibility. These results suggest that my measure can assist investors, analysts, or
researchers to compare earnings management flexibility across firms in the same
industry. This ability to compare one firm’s flexibility with that of another can come in
handy in evaluating the quality of a firm’s financial reports, stock picking or credit
granting decisions.
(7) (8)
(1) (2) (3) (4) (5) (6) Industry-adjusted CS
Barton and Simko’s measure Kasznik’s measure CS operating cycle operating cycle
Predicted Coefficient Marginal Coefficient Marginal Coefficient Marginal Coefficient Marginal
sign (SE) effects (SE) effects (SE) effects (SE) effects
lnoa þ 20.000 2 0.000

(0.000)
DTotal 0.000 0.000
accrual þ
Op_flex 2 (0.000) 20.292 * * * 2 0.105
(0.040)
Ind_Op_flex 2 2 0.204 * * * 2 0.073
(0.041)
DGDP 2 20.014 * * * 2 0.005 2 0.014 * * * 2 0.005 20.016 * * * 2 0.006 2 0.015 * * * 2 0.006
(0.003) (0.003) (0.003) (0.003)
DCPI þ 0.094 * * * 0.034 0.094 * * * 0.034 0.094 * * * 0.034 0.094 * * * 0.034
(0.005) (0.005) (0.005) (0.005)
lpb 2 20.023 * * * 2 0.008 2 0.023 * * * 2 0.008 20.023 * * * 2 0.008 2 0.023 * * * 2 0.008
(0.001) (0.001) (0.001) (0.001)
Pre_surp 2 22.054 * * * 2 0.737 2 2.054 * * * 2 0.737 22.061 * * * 2 0.739 2 2.056 * * * 2 0.737
(0.103) (0.103) (0.103) (0.103)
Exp_sur 2 20.005 2 0.002 2 0.005 2 0.002 20.005 2 0.002 2 0.005 2 0.002
(0.010) (0.010) (0.010) (0.010)
LROA 2 21.353 * * * 2 0.485 2 1.351 * * * 2 0.485 21.414 * * * 2 0.507 2 1.388 * * * 2 0.498
(0.125) (0.125) (0.126) (0.125)
L2ROA 2 20.585 * * * 2 0.210 2 0.584 * * * 2 0.210 20.681 * * * 2 0.244 2 0.647 * * * 2 0.232
(0.129) (0.129) (0.131) (0.130)
L3ROA 2 20.241 * 2 0.087 2 0.240 * 2 0.086 20.216 2 0.078 2 0.220 * 2 0.079
(0.133) (0.133) (0.133) (0.133)
L4ROA 2 20.485 * * * 2 0.174 2 0.483 * * * 2 0.173 20.496 * * * 2 0.178 2 0.487 * * * * 2 0.175
(0.132) (0.132) (0.133) (0.132)
L5ROA 2 0.156 0.056 0.156 0.056 0.166 0.060 0.166 0.059
(0.106) (0.106) (0.106) (0.106)
(continued)
management
Earnings
control for BigN auditors

flexibility

Table VII.
369
11,4
370
RAF
Table VII.
(7) (8)
(1) (2) (3) (4) (5) (6) Industry-adjusted CS
Barton and Simko’s measure Kasznik’s measure CS operating cycle operating cycle
Predicted Coefficient Marginal Coefficient Marginal Coefficient Marginal Coefficient Marginal
sign (SE) effects (SE) effects (SE) effects (SE) effects
L6ROA 2 0.042 0.015 0.043 0.015 0.053 0.019 0.052 0.019

(0.094) (0.094) (0.094) (0.094)
L7ROA 2 0.076 0.027 0.076 0.027 0.078 0.028 0.080 0.029
(0.091) (0.091) (0.091) (0.091)
L8ROA 2 0.071 0.026 0.067 0.024 0.063 0.023 0.068 0.024
(0.089) (0.089) (0.089) (0.089)
BigN 20.220 * * * 2 0.082 2 0.221 * * * 2 0.082 20.221 * * * 2 0.082 2 0.219 * * * 2 0.081
(0.017) (0.017) (0.017) (0.017)
Intercept 20.386 * * * 2 0.386 * * * 20.343 * * * 2 0.385 * * *
(0.020) (0.020) (0.021) (0.020)
Observations 70,349 70,349 70,349 70,349
McFadden R 2 0.0195 0.0195 0.0201 0.0198
Notes: Significant at: *10, * *5 and * * *1 percent levels (two-tailed); this table examines the relation between the probability of missing analyst’s
forecasts and earnings management flexibility, after controlling for potential difference between Big 4 vs non-Big 4 clients; BigN is a dummy variable that
takes on a value of 1 if the firm’s auditor is one of the big 8/6/5/4 accounting firms and 0 otherwise
One limitation of this study is that I focus on a measure of the upper limit on a firm’s Earnings
earnings management flexibility. In addition to a cap on the discretionary accruals a management
firm can incur, a firm also is likely to be constrained on the negative discretionary
accruals it can incur (i.e. a floor on the discretionary accruals or a cap on the cookie jar flexibility
reserve). Although this lower limit on earnings management may not be as stringent as
the upper limit, it can have an impact on a firm’s reporting strategy and performance.
Moreover, similar to other empirical studies, the proxies used in this study suffer from 371
measurement errors and can introduce noise to the results of my analyses.
Notes
1. The costs of violation include the cost of being caught by auditors and regulators. If caught,
both the manager and the firm face legal costs, damage to reputation, significant drop in
stock price around the announcement of SEC investigation and restatement, etc.
(Dechow et al., 1996; Palmrose et al., 2004; McNamee et al., 2000; Levitt, 2000).
2. We omit fraud cases and limit out discussion to earnings management within the constraints
of GAAP. In other words, if a manager cannot attain his target earnings with accruals from
the allowable set, I assume that he does not have the flexibility to manage earnings.
3. If a firm has unreversed negative accruals (i.e. B_Accrut , 0), then the level of Accrut it can
incur increases (i.e. greater earnings management flexibility). By taking negative
discretionary accruals in one quarter, a firm can build up its cookie jar reserve
of discretionary accruals for future periods. There is also a limit on the level of negative
discretionary accruals a firm can take in any one quarter without attracting the attention of
auditors and/or regulators. The restriction on positive vs negative discretionary accruals
may not be the same though. Further, it is possible that managers, observing a high level of
B_Accrut, change the governance structure or switch auditors to expand the flexibility limit.
However, this expansion of flexibility limits is also constrained and costly.
4. One limitation of the proposed measure is that it captures the constraint on the manager’s
ability to manage earnings through accruals only. The manager can still engage in real
activities, such as accelerating sales through increased price discounts, to inflate earnings
(Roychowdhury, 2006) even though he runs out of the flexibility to manage earnings via
accruals. However, earnings management through real activities is costly and also subject to
constraints.
5. Hribar (2000) documents that changes in the working capital account make up 46 percent of
total accruals and depreciation accounts for another 28 percent. Assuming that manipulation
of depreciation is not as flexible as manipulation of accounts receivable and inventory
accounts, then it is reasonable to assume that more than 46 percent of the discretionary
accruals are from the working capital accounts.
6. The average operating cycle of a firm in the prior 20 quarters has also been used as a proxy
for the reversal rate and the results are qualitatively the same.
7. Two alternative proxies for the reversal rate of accruals are also considered: (i) the reversal
rate of working capital estimated using Dechow and Dichev’s (2002) model. Coefficients on
lagged cash flows are examined. The index i of the longest lead cash flows that have a
significantly positive coefficient is assumed to represent the average reversal rate of the
firm’s working capital accruals. For example, if the coefficients of CFOtþ 1, CFOtþ2 are
significant but that of CFOtþ3 is not, then the reversal rate is three quarters. The reversal
rate estimated by the auto-regression of total accruals. The coefficients of the lagged total
accruals are examined to determine the reversal rate as in (i). Results using these alternative
proxies of reversal rates are similar to those reported in the tables.
RAF 8. This method of computing unreversed discretionary accruals for a quarter likely overstates
the constraint placed by discretionary accruals incurred in prior periods (assuming positive
11,4 discretionary accruals) by assuming that accruals all reverse at the end of the operating
cycle and hence limit the flexibility available throughout the whole operating cycle. Accruals
incurred in prior periods likely reverse throughout the whole operating cycle and hence the
constraint they place on flexibility available likely decrease over time. However, the reversal
can take various patterns, such as evenly throughout the cycle, or reverse at an accelerated
372 rate at the beginning of the cycle and gradually slow down till the end of the cycle, etc. By
assuming all accruals reverse at the end of the cycle, I generate a more conservative estimate
of flexibility available.
9. A sensitivity test has also been performed using the maximum CLDA in the
industry-quarter as a proxy for the upper flexibility limit and the results are similar to
those reported in Section 6.
10. Hribar and Collins (2002) document that accruals calculated by the balance-sheet approach
can suffer from the articulation problem and advocate the use of the cash-flow-statement
approach to calculate the accruals. The earliest date accruals data are available is 1987.
11. One advantage of using the consensus analysts’ forecast, rather than earnings of the same
quarter last year, as a benchmark is that the consensus analysts’ forecast likely has
accounted for the reversal effect of prior quarter accruals. Thus, the earnings surprise
calculated using the analysts’ forecast would not have a mechanical relation with the
cumulative discretionary accruals flexibility measure.
12. I also perform the analysis using consensus forecasts that exclude analysts’ forecasts issued
90 days before the earnings announcement (i.e. stale forecasts) in constructing miss. I further
exclude analysts’ forecasts issued before earnings announcement of the current quarter in
constructing Exp_sur. The results are insensitive to these changes.
13. This proxy for pre-managed earnings change does not account for the effect of management
through real activities.
14. The negative effect of Pre_surpt can also be caused by a mechanical relationship. Since the
probability of missing analysts’ forecasts decreases in current reported earnings (REt) while
Pre_surpt increases in REt, this can lead to the observed negative coefficient of Pre_surpt.
15. Alternatively, I perform the probit analysis using firms with absolute forecast errors less
than 10 cents, 5 cents, 2 cents, etc. Results (untabulated) are similar to those reported in
Table III.
16. I also perform the analysis using flexibility measures constructed with discretionary
accruals estimated by the performance-matched model (Kothari et al., 2005). Firms with
poor performance in the prior quarters are more likely to have managed earnings
upwards, and hence, have low flexibility in the current quarter. Using
performance-matched samples to estimate discretionary accruals likely removes most of
the variation in my flexibility measure. As expected, the correlation between flexibility
and probability of missing analysts’ forecasts is weaker, but the coefficient of the
flexibility measure continues to be negative and significant. Accordingly, a firm’s
flexibility has incremental effect on its probability of missing analyst’s forecast after
controlling for its lagged performance.
17. By restricting the sample to the Barton and Simko sample period, I can compare the results
using the industry-adjusted measure with those using the raw measure in Table III
Panel A. Any difference in the two sets of results cannot be caused by the difference in
sample construction but rather by the industry adjustment.
18. The flexibility measure is converted into dollars per share, by multiplying the measure by Earnings
the amount of total assets and then divided by the number of shares outstanding.
The pre-managed earnings per share is then subtracted from this converted flexibility management
measure to construct Pmeet. flexibility
References
Abarbanell, J. and Lehavy, R. (2003), “Biased forecasts or biased earnings? The role of reported 373
earnings in explaining apparent bias and over/underreaction in analysts’ earnings
forecasts”, Journal of Accounting and Economics, Vol. 36 Nos 1-3, pp. 105-46.
Anthony, J. and Ramesh, K. (1992), “Association between accounting performance measures and
stock prices: a test of the life cycle hypothesis”, Journal of Accounting & Economics, Vol. 15
Nos 2/3, pp. 203-27.
Barton, J. and Simko, P. (2002), “The balance sheet as an earnings management constraint”,
The Accounting Review, Vol. 77, pp. 1-27.
Bartov, E. and Cohen, D. (2008), “The “numbers game” in the pre- and post-Sarbanes-Oxley eras”,
working paper, New York University, New York, NY.
Burgstahler, D. and Dichev, I. (1997), “Earnings management to avoid earnings decreases and
losses”, Journal of Accounting & Economics, Vol. 24 No. 1, pp. 99-126.
Business Wire (2002), “Law offices of Marc S. Henzel announces class action lawsuit against
Amdocs Limited”, 2 August.
Choy, H.L. (2010), “Impact of the Sarbanes-Oxley Act on firms’ earnings management flexibility”,
working paper, Drexel University, Philadelphia, PA.
Cohen, D., Dey, A. and Lys, T. (2008), “Real and accrual based earnings management in the
pre and post Sarbanes Oxley periods”, The Accounting Review, Vol. 83 No. 3, pp. 757-87.
DeFond, M. (2002), “Discussion of the balance sheet as an earnings management constraint”,
The Accounting Review, Vol. 77, pp. 29-33.
DeFond, M. and Jiambalvo, J. (1994), “Debt covenant violation and manipulation of accruals”,
Journal of Accounting & Economics, Vol. 17 Nos 1/2, pp. 145-76.
DeFond, M. and Park, C. (1997), “Smoothing income in anticipation of future earnings”, Journal of
Accounting & Economics, Vol. 23 No. 2, pp. 115-39.
Dechow, P. (1994), “Accounting earnings and cash flows as measures of firm performance.
The role of accounting accruals”, Journal of Accounting & Economics, Vol. 18 No. 1,
pp. 3-42.
Dechow, P., Sloan, R. and Sweeney, A. (1996), “Causes and consequences of earnings
manipulation: an analysis of firms subject to enforcement actions by the SEC”,
Contemporary Accounting Research, Vol. 13 No. 1, pp. 1-36.
Degeorge, F., Patel, J. and Zeckhauser, R. (1999), “Earnings management to exceed thresholds”,
The Journal of Business, Vol. 72 No. 1, pp. 1-33.
Elgers, P., Pfeiffer, R. and Porter, S. (2003), “Anticipatory income smoothing: a re-examination”,
Journal of Accounting and Economics, Vol. 35 No. 3, pp. 405-22.
Erickson, M. and Wang, S. (1999), “Earnings management by acquiring firms in stock for stock
merger”, Journal of Accounting & Economics, Vol. 27 No. 2, pp. 149-76.
Fields, T., Lys, T. and Vincent, L. (2001), “Empirical research on accounting choice”, Journal of
Accounting & Economics, Vol. 31 Nos 1-3, pp. 255-307.
Francis, J. (1984), “The effect of audit firm size on audit prices: a study of the Australian market”,
Journal of Accounting & Economics, Vol. 6 No. 2, pp. 133-51.
RAF Francis, J. and Simon, D. (1987), “A test of audit pricing in the small-client segment of the US
audit market”, The Accounting Review, Vol. 62 No. 1, pp. 145-57.
11,4
Francis, J. and Stokes, D. (1986), “Audit prices, product differentiation and scale economies:
further evidence from the Australian market”, Journal of Accounting Research, Vol. 24
No. 2, pp. 383-93.
Francis, J., Maydew, E. and Sparks, H. (1999), “The role of Big 6 auditors in the credible reporting
374 of accruals”, Auditing: A Journal of Practice & Theory, Vol. 18 No. 2, pp. 17-34.
Fudenberg, D. and Tirole, J. (1995), “A theory of income and dividend smoothing based on
incumbency rents”, Journal of Political Economy, Vol. 103 No. 1, pp. 75-93.
Healy, P. (1985), “The effect of bonus schemes on accounting decisions”, Journal of Accounting
& Economics, Vol. 7 Nos 1-3, pp. 85-113.

Hribar, P. (2000), “The market pricing of components of accruals”, working paper,
Cornell University.
Hribar, P. and Collins, D. (2002), “Errors in estimating accruals: implications for empirical
research”, Journal of Accounting & Economics, Vol. 40 No. 1, pp. 105-34.
Kasznik, R. (1999), “On the association between voluntary disclosure and earnings
management”, Journal of Accounting Research, Vol. 31 No. 1, pp. 57-81.
Kothari, S.P., Leone, A. and Wasley, C. (2005), “Performance matched discretionary accrual
measures”, Journal of Accounting and Economics, Vol. 39 No. 1, pp. 163-97.
Leone, A. and Rock, S. (2002), “Empirical tests of budget ratcheting and its effect of managers’
discretionary accrual choices”, Journal of Accounting & Economics, Vol. 33 No. 1, pp. 43-67.
Lev, B. and Thiagarajan, R. (1993), “Fundamental information analysis”, Journal of Accounting
Research, Vol. 31 No. 2, pp. 190-215.
Levitt, A. (2000), “Testimony concerning Commission’s auditor independence proposal before the
Senate Subcommittee on Securities Committee on Banking, Housing, and Urban Affairs”,
September 28, available at: www.sec.gov/news/testmony/ts152000.htm
McCulloch, B. (1998), “Multi-period incentives and alternative dials for earnings management”,
working paper, New Zealand Treasury.
McNamee, M., Dwyer, C., Schmitt, C.H. and Lavelle, L. (2000), “Accounting wars: powerful
auditor-consultants are the targets of Arthur Levitt’s crusade”, Business Week Online,
September, p. 25.
Matsumoto, D. (2002), “Management’s incentives to avoid negative earnings surprises”,
The Accounting Review, Vol. 77 No. 3, pp. 483-514.
Murphy, K. (2000), “Performance standards in incentive contracts”, Journal of Accounting
& Economics, Vol. 30 No. 3, pp. 245-78.
Palmrose, Z.V. (1986), “Audit fees and auditor size: further evidence”, Journal of Accounting
Research, Vol. 24 No. 1, pp. 97-110.
Palmrose, Z.V., Richardson, V. and Scholz, S. (2004), “Determinants of market reactions to
restatement announcements”, Journal of Accounting and Economics, Vol. 37 No. 1,
pp. 59-89.
Roychowdhury, S. (2006), “Earnings management through real activities manipulation”, Journal
of Accounting and Economics, Vol. 42 No. 3, pp. 335-70.
Skinner, D. and Sloan, R. (2002), “Earnings surprises, growth expectations, and stock returns or
don’t let an earnings torpedo sink your portfolio”, Review of Accounting Studies, Vol. 7
Nos 2/3, pp. 289-312.
Sweeney, A.P. (1994), “Debt-covenant violations and managers’ accounting responses”, Journal Earnings
of Accounting & Economics, Vol. 17 No. 3, pp. 281-308.
Teoh, S.H., Welch, I. and Wong, T.J. (1998a), “Earnings management and the long-run market
management
performance of initial public offerings”, The Journal of Finance, Vol. 53 No. 6, pp. 1935-74. flexibility
Teoh, S.H., Welch, I. and Wong, T.J. (1998b), “Earnings management and the underperformance
of seasoned equity offerings”, Journal of Financial Economics, Vol. 50 No. 1, pp. 63-99.
Watts, R. and Zimmerman, J. (1986), Positive Accounting Theory, Prentice-Hall, 375
Englewood Cliffs, NJ.
Further reading
Bureau of Economic Analysis (2002), National Accounts Data, available at: www.bea.doc.gov
Bureau of Labor Statistics (2002), Consumer Price Indexes Program, available at: www.bls.gov/
home.htm
Dechow, P. and Dichev, I. (2002), “The quality of accruals and earnings: the role of accrual
estimation errors”, The Accounting Review, Vol. 77, pp. 35-59.
Maddala, G.S. (1983), Limited-dependent and Qualitative Variables in Econometrics,
Cambridge University Press, Cambridge.
Statement on Auditing Standards 56 (1988), Analytical Procedures, April.
About the author

Hiu Lam Choy is an Assistant Professor in the LeBow College of Business at Drexel University.
Her research interests include auditing, earnings management, executive compensation, and
corporate governance. Hiu Lam Choy can be contacted at: [email protected]
(The Appendix follows overleaf.)
To purchase reprints of this article please e-mail: [email protected]

Or visit our web site for further details: www.emeraldinsight.com/reprints
11,4
376
RAF
Table AI.
Definition of variables
used in the empirical tests
Variables Definition
missit Equals 1 if firm i misses the mean consensus analysts’ forecast in quarter t and 0 otherwise. A firm misses the consensus analysts’ forecast if the difference
Appendix
between the actual earnings per share and the last consensus analysts’ forecast available before earnings announcement
is negative
Flexit Flexibility proxy. Three different proxies are used in the test:
(i) Barton and Simko’s measure – lagged net operating asset flexibility deflated by sales (lnoa). This flexibility is constructed as in Barton and Simko (2002).
Net operating assets are defined as shareholders’ equity less cash and marketable securities, plus debt. The net operating assets are then deflated by sales to
get lnoa. The higher the net operating assets deflated by sales, the lower the flexibility
(ii) Kasznik’s measure – change in total accruals (DTotal accrual). This flexibility measure is computed as the difference between total accruals of this
quarter and those of the same quarter last year, divided by total accruals of the same quarter last year. Total accruals are computed as the difference between
income before extraordinary items and cash flow from operations
Operating cycle flexibility (Op_flex), which is calculated as the upper
flexibility bound less the sum of discretionary accruals cumulated over Qi quarters, where Qi is the number of quarters in the firm’s operating cycle.
The firm’s operating cycle is computed as:
Operating cycle ¼ þ 2 *90
Sales Cost of goods sold Cost of goods Sold
The operating cycle is computed cross-sectionally. Firms in the same two-digit SIC codes and the same quarter are grouped together to compute the average
operating cycle. This average operating cycle is used as a proxy for the reversal rate for all firms in that industry-quarter. Assuming that the operating cycle
is three quarters, the cumulative lagged discretionary accruals (CLDA) is calculated as: CLDAt ¼ DAt21 þ DAt22 þ DAt23
The upper flexibility limit is then calculated as the mean of the cumulative lagged discretionary accruals of the industry (defined by two-digit SIC code),
CLDA, plus two times the standard deviation of these lagged accruals (sðCLDAÞ). That is, Upper Flexibility Limit ¼ CLDA þ 2sðCLDAÞ:
The discretionary accruals used to construct Op_flex are estimated cross-sectionally by the residuals of:
TAt a DSalest PPE t
¼ þ b1 þ b2 þ b3 SGt þ b4 Dividt þ 1t ð2Þ
Pre_surpit Pre-managed earnings change of the current quarter. The pre-managed earnings of a quarter are defined as the reported earnings less the discretionary
accruals. The pre-managed earnings change is then calculated as the difference between the current pre-managed earnings and
the reported earnings of the same quarter last
year (i.e. (REt 2 DAt 2 REt2 4))
Exp_surit Expected future earnings change. It is computed by subtracting the current reported earnings stripped of discretionary accruals from the expected earnings
of the same quarter of the following year, t þ 4 [i.e. (E[TEtþ 4] 2 REt þ DAt)]. The expected earnings are proxied by the first consensus analysts’ forecast of
t þ 4 released after the earnings announcement of t
DGDPt Percentage change in real gross domestic product
DCPIt Percentage change in consumer price index
lpbit Lagged market-to-book of equity ratio
Note: This table provides a detailed description of the variables used in probit analyses in the paper
This article has been cited by:
1. LaitinenErkki Kalervo, Erkki Kalervo Laitinen. 2018. Implied expense theory in financial reporting: a
steady-state approach. Journal of Financial Reporting and Accounting 16:1, 49-83. [Abstract] [Full Text]
[PDF]
2. Camelia Vasilescu, Yuval Millo. 2016. Do industrial and geographic diversifications have different effects
on earnings management? Evidence from UK mergers and acquisitions. International Review of Financial
Analysis 46, 33-45. [Crossref]
3. MindakMary P., Mary P. Mindak, SenPradyot K., Pradyot K. Sen, StephanJens, Jens Stephan. 2016.
Beating threshold targets with earnings management. Review of Accounting and Finance 15:2, 198-221.
[Abstract] [Full Text] [PDF]

Review of Accounting and Finance: Article Information

Uploaded by

Copyright:

Available Formats

Review of Accounting and Finance: Article Information

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Review of Accounting and Finance: Article Information

Uploaded by

Copyright:

Available Formats

Review of Accounting and Finance

Assessing earnings management flexibility

Permanent link to this document:

*Related content and download information correct at time of download.

rate of these accruals.

JEL classification – M40, M41

allowable set of accruals[2] which I denote as the “earnings management flexibility

2. Effect of prior earnings management and reversal rate on current

Effect of prior earnings management practice

Effect of the reversal rate of accruals

Effect of the width of flexibility boundaries

TAt a DSalest PPE t

Avg A=R Avg Inventory Avg A=P

Proxy for earnings management flexibility used

Proxy for earnings management flexibility limits

4. Other determinants of costs of earnings management and flexibility

growth is measured as the annual percentage change in real GDP.

5. Sample selection and data description

(4) (5) (6) (7) (8) (9) (10) (11)

for sample firms

(4) (5) (6) (7) (8) (9) (10) (11)

Definition of SIC sectors

flexibility measure is consistent with these two flexibility measures proposed in

6. Flexibility and missing analysts’ forecasts

Managers consider expected future earnings when making earnings management

variables in the earnings

(1) (2) (3) (4) (5) (6)

Panel B: correlations among variables for the earnings surprise test

(1) (2) (3) (4) (5) (6)

Panel A: Barton and Simko’s sample period (1993-1999)

Probit analysis of the

DGDP 2 2 0.023 * * * 2 0.008 2 0.023 * * * 2 0.008 2 0.024 * * * 2 0.009

Flexibility measures examined in the same regression analysis

Industry-adjusted flexibility measures

lnoa þ 2 0.000 2 0.000 2 0.000 2 0.000

(1) (2) (7) (8) (9) (10)

7. Additional sensitivity analysis

Barton and Simko’s measure

(1) (2) (3) (4) (5) (6)

lnoa þ 0.000 0.000

Probit analysis of the

lnoa þ 20.000 2 0.000

control for BigN auditors

Probit analysis of the

L6ROA 2 0.042 0.015 0.043 0.015 0.053 0.019 0.052 0.019

& Economics, Vol. 7 Nos 1-3, pp. 85-113.

About the author

(The Appendix follows overleaf.)

To purchase reprints of this article please e-mail: [email protected]

You might also like