10 1016@j Irfa 2018 10 001
10 1016@j Irfa 2018 10 001
10 1016@j Irfa 2018 10 001
PII: S1057-5219(18)30153-4
DOI: doi:10.1016/j.irfa.2018.10.001
Reference: FINANA 1264
To appear in: International Review of Financial Analysis
Received date: 3 March 2018
Revised date: 2 October 2018
Accepted date: 3 October 2018
Please cite this article as: Basiem Al-Shattarat, Khaled Hussainey, Wasim Al-Shattarat ,
The impact of abnormal real earnings management to meet earnings benchmarks on future
operating performance. Finana (2018), doi:10.1016/j.irfa.2018.10.001
This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.
ACCEPTED MANUSCRIPT
Basiem Al-Shattarat
Corresponding author
Prince Sultan University
bshattarat@psu.edu.sa
P.O.Box No. 66833 Rafha Street, Riyadh 11586 Saudi Arabia
PT
Mobile +966(0)557282170
RI
Khaled Hussainey
University of Portsmouth, Portsmouth, UK
SC
Professor in Accounting and Financial Management
khaled.hussainey@port.ac.uk
Richmond Building, Portland Street, Portsmouth, PO1 3DE
NU
Mobile 07727190105 Office 02392844715
MA
Wasim Al-Shattarat
Gulf University for Science and Technology (GUST), Kuwait
Assistant Professor of Accounting
ED
Shattarat.W@gust.edu.kw
Block 5, Building 1
PT
1
ACCEPTED MANUSCRIPT
Abstract
PT
signalling perspectives), this study investigates the association between firms that manipulate
RI
their business operations to meet earnings benchmarks (i.e., zero earnings, last year’s
earnings) and subsequent operating performance. We examine the effects of the magnitude of
SC
real earnings management on firms’ future performance for the period 2009 to 2015 for UK
NU
firms. Our analysis shows that the manipulation of operating activities such as sales,
significantly positive consequence for firms’ subsequent operating performance and signals
firms’ good future performance. We also find that firms that manipulate their operating
ED
subsequent operating performance. The findings of this research lend support to our
PT
understanding of the process that management follows to evaluate costs and benefits of real
CE
earnings management.
2
ACCEPTED MANUSCRIPT
1. Introduction
Since reported earnings are the outcome of accounting choices and the underlying
business operations, firms may utilise alternative earnings management tools to manipulate
their earnings to achieve certain earnings benchmarks (Graham et al., 2005; Roychowdhury,
2006; Cohen and Zarowin, 2010; Gunny, 2010; Zang, 2012). The common tools of earnings
PT
(AEM) and real earnings management (REM). AEM takes place when managers control their
RI
reported earnings through exploitation of the accounting discretion associated with generally
SC
accepted accounting principles (GAAP). REM involves managers’ efforts to alter their
reported earnings by making suboptimal decisions on the timing and scales of underlying
NU
business activities (Roychowdhury, 2006). While AEM has no direct cash flow consequences
and is therefore less likely to destroy long-term firm value (Dechow et al., 2010), REM
MA
influences operations with direct effects on cash flows. Therefore, to achieve a good stock
market performance and financial position, managers are incentivised to manage earnings
ED
based on accounting numbers not only to maximise the value of the firms but also to extract
PT
In this study, we examine the relationships between REM and firms’ subsequent
CE
operating performance. Specifically, we investigate whether United Kingdom (UK) firms that
manipulate their sales, discretionary expenses and production around zero earnings and last
AC
year’s earnings to report higher earnings realise an impact from these activities on future
identify firms that are more likely to manage earnings upward. Specifically, we achieve this
based on the firms’ ability to meet (1) zero earnings, and (2) last year’s earnings.1
1
Due to data access limitations, this study does not explore the third benchmark, “analysts’ forecast”.
3
ACCEPTED MANUSCRIPT
earnings benchmarks, e.g., reporting positive profit, avoiding earnings decrease and avoiding
negative earnings surprises, are stronger around firms that are more likely to manipulate their
management and different motivations. In turn, this association may influence the earnings
management choices of firms. (Degeorge, et al., 1999; Dechow and Dichev, 2002; Graham et
PT
al., 2005; Roychowdhury, 2006). Burgstahler and Dichev (1997) explain that, based on
RI
transaction and information costs, investors derive economic decisions from heuristics or
reference points, such as zero level or earnings changes, as well on the ‘surprises’ that zero
SC
earnings generate. In this way, a loss or decrease in earnings may send a negative signal to
NU
outsiders, particularly credit rating agencies and stock analysts who assess the firm. This
signal, in turn, negatively affects a firm’s credit ratings and the costs of the debt. However,
MA
outsiders may attach different weights to such a signal, depending on the firm’s previous
signals.
ED
REM. One view is the ‘opportunistic earnings management’ argument that managers who use
PT
REM deviations from normal business strategy to manage reported earnings opportunistically
CE
mislead outside investors on their assessments of firms’ performance, thus potentially leading
to a decline in subsequent performance. Consistent with this view, the prior literature has
AC
documented that firms that engage in REM experience a negative impact on subsequent
financial performance and firm value (Bens et al., 2002). Moreover, prior research observes a
decline in future performance among firms that engage in higher REM to meet certain
financial reporting benchmarks (Ewert and Wagenhofer, 2005; Graham et al., 2005;
Roychowdhury, 2006; Cohen and Zarowin, 2010; De Jong et al., 2014; Alhadab et al., 2015;
4
ACCEPTED MANUSCRIPT
The opposite view is the ‘signalling earnings management’ argument, which claims that
firms utilise REM to signal good future performance and distinguish themselves from poor
performance (Roychowdhury, 2006). REM may not necessarily result in a more significant
activities are less likely to significantly affect the operations of firms that occupy strong
financial and market positions and intend to use earnings to communicate favourable private
PT
information about future performance (Zang, 2012). Consistent with this view, prior research
RI
observed a positive impact on the client’s cash flow and good future performance among
SC
firms that manage earnings upward by utilising more REM to meet/beat earnings benchmarks
(Bartov et al., 2002; Gunny, 2010; Taylor and Xu, 2010; Zhao et al., 2012).
NU
This study contributes to the accounting literature in a number of ways. First, it
contributes to the extant empirical research on the relationship between REM and future
MA
performance (e.g., Taylor and Xu, 2010; Kothari et al., 2016; Leggett et al., 2016) by
providing empirical evidence on the relationships between the three types of REM to meet
ED
zero earnings and last year’s earnings and future operating performance in UK-listed firms.
To the best of our knowledge, there is limited evidence on the relationship between REM and
PT
future performance around firms that meet zero earnings and last year’s earnings (Gunny,
CE
2010; Zhao et al., 2012; Francis et al., 2016a). The studies in this field are US-based, and
Second, this study is the first to use the absolute value of each measure to capture the
general level of REM activities on future operating performance. For example, Gunny (2010)
examines a United States (US) sample and uses REM activities as the indicator variable equal
to one if the residual from Research and Development (R&D), Selling, General, and
Administrative expenses (SG&A), and production models is in the lowest (highest) quintile
and zero otherwise. In contrast to previous research (e.g., Gunny, 2010), we consider how to
5
ACCEPTED MANUSCRIPT
avoid the drawbacks of binning continuous variables. That is, we assess the potential loss of
power and loss of precise average effects that could arise by estimating the mean effect of the
measures in the upper quantile rather than estimating the means effect of all observations
(Harrell, 2015).
focuses on cross-sectional dependence (e.g., Gunny, 2010; Zhao et al., 2012) but does not
PT
examine the issues created by the presence of both cross-sectional and time-series
RI
dependence. In this study, we use the Newey-West standard error-corrected Fama-MacBeth
SC
procedure as a method that corrects for dependence in one direction and typically assumes
independence in the other2. This therefore adds value to the robustness of the results by
NU
correcting for potential bias and inconsistency issues in the estimates and overcoming
heteroscedasticity problems.
MA
In the UK context, the accounting standards setting does differ from that in the US,
which could affect the inferences drawn from this research (e.g., the mandatory adoption of
ED
International Financial Reporting Standards (IFRS) in the UK). If mandatory IFRS adoption
has an influence, either positive or negative, on AEM, it may also have an influence on REM.
PT
However, Cohen et al. (2008) and Zang (2012) provide evidence that the presence of tighter
CE
accounting standards and less accounting flexibility leads managers to substitute AEM with
REM, clarifying that REM can occur independent of manipulation through AEM. It is,
AC
however, more difficult to track REM for outsiders as it can be masked in the form of
everyday business transactions, by involving, for example, decisions about changes in the
timing or structuring of a transaction (Cohen and Zarowin, 2010). Under IFRS, for instance,
research and advertising costs are expensed in the period in which they are incurred.
Therefore, reducing these costs reported affects income. Moreover, developments costs are,
2
Our results remain qualitatively unchanged when we run an ordinary least squares model with robust standard
errors adjusted for heteroskedasticity and clustered by firm or a firm-fixed effect model instead of a Fama-
McBeth model.
6
ACCEPTED MANUSCRIPT
in the first instance, expensed rather than capitalised due to uncertainty issues regarding the
In addition, there are numerous differences associated with institutional and capital
market characteristics between countries. US capital markets are much larger than those in
the UK and are arguably subject to more regulatory scrutiny. Ball et al. (2000) recognise that
PT
the UK has the least regulated accounting, least regulated litigations and least issuance of
RI
public debt. Moreover, Brown and Higgins (2001) find evidence that UK firms have smaller
SC
holdings of stock than their US counterparts do and thus suggest that managers in those UK
firms have fewer incentives to manage earnings to avoid reporting bad news. In addition,
NU
recent literature has shown that differences in the expectations of management behaviour in
different countries may explain the differences in the level of earnings management practices
MA
across countries (Leuz et al., 2003; Brown and Higgins, 2005; Han et al., 2010; Francis et al.,
2016b). Thus, it is not clear that UK firms have the same incentives to meet earnings
ED
While the previous empirical results are mixed, our findings show that UK firms
PT
utilising REM to meet earnings benchmarks do not necessarily have significantly negative
CE
consequences for firms’ subsequent operations. Therefore, our findings are in line with the
signalling earnings management argument in which firms that meet earnings benchmarks
AC
utilise REM activities to convey their private information to signal their future good
performance and distinguish themselves from poor performance; this subsequently enhances
investors’ ability to predict firms’ performance. In the absence of meeting earnings, the
results also support the opportunistic earnings management argument. Therefore, investors
Our findings have two important implications for both standard setting and stock
7
ACCEPTED MANUSCRIPT
market participants. First, it informs regulators about how managers use accounting judgment
using REM to meet earnings benchmarks and to make financial statements more informative
for investors and financial analysts. This, as argued by Healy and Wahlen (1999, p. 369),
“can arise if certain accounting choices or estimates are perceived to be credible signals of a
firm’s financial performance”. In addition, regulators and stock market authorities may
consider those actions that deviate from normal operational business activities to reduce the
PT
scope of earnings management by increasing levels of corporate disclosure and enhancing the
RI
quality of corporate reporting. Second, by facilitating a better understanding of REM on
firms’ future performance, stock market participants (e.g. investors and financial analysts)
SC
may consider the consequences of REM activities as well as AEM activities when they
NU
making investment decisions.
The reminder of the study is organised as follows. Section 2 reviews the literature and
MA
develops our hypotheses. Section 3 discusses the sample, data, and research design. Section 4
Since all REM activities lead to higher net income in the current period, these activities
will inevitably have consequences. However, the empirical results are mixed. Hence, the
CE
performance have attracted opposing explanations for why the managers adopt REM.
The extant literature shows that when a firm’s REM manages-up earnings, this reduces
the firm’s value, which will harm the firm’s future performance (value destroying). In the
absence of meeting earnings benchmarks, Bens et al. (2002) find that firms that manipulate
earnings shift capital away from real investment to stock repurchases by reducing the R&D
experience a marginally negative impact on future operating performance. Other studies also
8
ACCEPTED MANUSCRIPT
find that firms — whether they engage in REM activities — with high net operating assets
Simko, 2002; Hirshleifer et al., 2004). In a recent study, Mizik (2010) finds that firms that
manage earnings upward by engaging in myopic REM activities through reducing marketing
and R&D spending experience a greater negative impact on future stock returns and future
financial performance. Recently, Vorst (2016) shows that, on average, firm-years with real
PT
activity manipulation (e.g., a reversal of an abnormal cut in discretionary investment) are
RI
associated with lower long-term operating performance. However, he finds that such results
SC
vary significantly depending on the various incentives offered to engage in REM, as well as
that chief financial officers (CFOs) are willing to manipulate REM activities to meet
MA
analysts’ earnings expectations, even if such manipulations would decrease long-term firm
value.3 The surveyed chief executive officers and CFOs acknowledge that they face a trade-
ED
off between meeting short-term earnings targets and making long-term optimal business
decisions.4 Furthermore, they argue that 80% of the participants wish to adopt REM to
PT
achieve short-term earnings targets. On the extent to which managers sacrifice real resources
CE
to manage earnings, Graham et al. (2005) show that more than 55% of the managers forfeit
positive net present value investment projects to meet analysts’ earnings forecasts. Other
AC
research on myopic behaviour and REM confirms Graham and colleagues’ survey evidence
which is costly and directly harmful to a firm’s future operation (Baber et al., 1991; Bhojraj
3
Graham et al. (2005) find that CFOs prefer to manage earnings via economic actions such as postponing or
eliminate hiring, R&D, advertising, or even investment rather than within-GAAP accounting choices.
4
The primary incentives for managers to meet short-term objectives are stock prices and career and reputation
concerns (Graham et al., 2005).
9
ACCEPTED MANUSCRIPT
Moreover, Zhang (2008) evinces that US firms that meet analysts’ cash flow forecast
engage in higher REM activities through discretionary expenditures, production and sales to
inflate earnings. Additionally, they find that REM firms experience deterioration in
subsequent operating performance. Furthermore, Bhojraj et al. (2009) show that firms that
beat analysts’ forecasts have negative operating future performance and stock market
performance in the subsequent three years. In a similar study to that of Zhang (2008), Leggett
PT
et al. (2016) show that firms that engage in REM through discretionary expenditures are
RI
negatively associated with lower subsequent future performance in terms of both return on
SC
assets and cash flows from operations than non-REM firms meeting/beating earnings
benchmarks. However, the notion that REM is value destroying is consistent with investor
NU
perceptions. In a recent survey, De Jong et al. (2014) find that analysts perceive that meeting
earnings benchmarks and smoothing earnings enhances investors’ perceptions of firm future
MA
performance and that all earnings management actions to meet earnings benchmarks,
Focusing on REM to just meet/beat zero earnings and last year’s earnings, Francis et
al. (2016a) examine whether firms that utilised REM to manage earnings upward are
PT
associated with the subsequent stock price risk, which is due to stocks being mispriced under
CE
REM. They find that prior REM has a positive association with stock price crashes in the
subsequent period. This result suggests that managers utilised REM activities to hide negative
AC
information but not positive information. In addition, they find that the impact of REM on the
stock return-crash risk increases after the Sarbanes-Oxley Act of 2002. By examining a
sample of all of California’s non-profit hospitals, Eldenburg et al. (2011) report a negative
relationship between REM and subsequent operating performance with managers whose
10
ACCEPTED MANUSCRIPT
Given that previous research on the consequences of REM pays little attention to the
bond market, Ge and Kim (2014) find that REM activities through overproduction causes
credit ratings to decline, and their results also show that overproduction and sales-based
manipulation increase the cost of borrowing money from the bond market. Similarly, Kim
and Sohn (2013) report a positive association between US firms in which utilised REM
meet/beat an earnings target and the implied cost of equity even after controlling for the
PT
effects of AEM.
RI
Apart from future performance, several studies examine other effects of real activity
SC
manipulation to maintain high stock prices (e.g., equity-offering firms). For instance,
Wongsunwai (2013) finds that IPO firms manage earnings around the IPO year and that IPO
NU
firms backed by higher-quality venture capitalists generally exhibit higher performance; thus,
they have lower real and accrual-based manipulation on average. Similarly, Alhadab et
MA
al. (2015) show that UK IPO firms that manage up earnings during the IPO year, either
through REM activities or AEM, have a higher probability of IPO failure and lower survival
ED
rates in subsequent periods. In addition, Cohen and Zarowin (2010) and Kothari et al. (2016)
provide evidence that, at the time of SEO, a firm that engages in income-increasing REM has
PT
a more negative future operating performance in the post-offering period than other firms that
CE
engage in AEM.
AC
REM may not necessarily have a significantly negative effect on firms’ future
operations, signalling an argument that claims that managers have better information about
firms’ future market and growth potential. They engage in REM because it is a way to signal
firms’ future value. The findings in the existing research present different views on the
relationship between REM activities and future operating performance. Focusing on two
common earnings benchmarks (i.e., zero earnings and last year’s earnings), Gunny (2010)
11
ACCEPTED MANUSCRIPT
examines the association between income-increasing REM and future performance.5 She
finds evidence that US firms that manage earnings upward to meet/beat earnings benchmarks
achieve a more positive impact on the client’s cash flow and subsequent operating
performance than other firms in the absence of just meeting/beating earnings benchmarks or
managing earnings upward through REM. This supports the signalling argument that firms
PT
Zhao et al. (2012) support Gunny’s (2010) findings; they find evidence that managers
RI
are generally willing to manage earnings upward through REM to meet/beat earnings
SC
benchmarks. In addition, they find a negative relationship between the abnormal REM and
Taylor and Xu (2010) provide evidence that US firms that meet/beat zero earnings and
analysts’ forecast with high abnormal production costs or/and low abnormal discretionary
ED
operating performance than control firms do when matched by industry, year and abnormal
PT
AEM. Their findings are consistent with those of Gunny (2010) and Zhao et al. (2012), which
CE
suggest that used-only REM offers more positive signalling effects about future firms’
performance than firms that used only AEM. Moreover, previous studies have observed good
AC
future performance among firms, which manipulated earnings to meet/beat the analysts’
5
Gunny (2010) uses indicator variables and classifies REM firms in the most extreme quintile of abnormal
REM activities.
6
Tan and Jamal (2006) suggest that managers manipulate operating activities by reducing the level of
accounting discretion (e.g., reduce their investments in R&D and increase those in advertising) to communicate
their firm’s superior earnings prospects to investors, and they attain costs and benefits of REM that allow the
firm to perform better in future performance (signalling mechanism). In addition, managers may also manipulate
earnings to extract personal benefits.
12
ACCEPTED MANUSCRIPT
Under the agency theory perspectives, the net effect of earnings management on a
firm’s value depends on whether managers manipulate earnings mainly to serve their own
interest (opportunistic earnings management) and thus mislead investors on their assessments
of firms’ performance (Healy and Palepu, 1993; Subramanyam, 1996; DeFond and Park,
1997). On the other hand, due to information asymmetry, investors usually do not have as
PT
much information as the managers. Therefore, managers may use earnings to communicate
RI
their private information on firms’ future performance and thus improve earnings’
SC
1998; Kothari, 2001; Sankar and Subramanyam, 2001). In addition, if managers sacrifice
NU
short-term value to manipulate earnings to signal their firms’ future performance (signalling
earnings management) and the market recognises the information in the signal, the benefits
MA
may offset the costs and eventually increase a firm’s value, thus protecting its long-term
value.
ED
performance have generated opposing explanations regarding the reasons the managers adopt
PT
REM. A negative correlation is found between REM activities and future performance of
CE
firms, which suggests that an opportunistic mechanism may affect the assumption of REM
(e.g., Ewert and Wagenhofer, 2005; Graham et al., 2005; Cohen et al., 2008; Cohen and
AC
Zarowin, 2010; Mizik, 2010; Zang, 2012; De Jong et al., 2014; Francis et al., 2016a; Kothari
et al., 2016; Leggett et al., 2016; Vorst, 2016), while a positive correlation between REM and
future performance of firms suggests that the signalling mechanism may affect the adoption
of REM (e.g., Gunny, 2010; Taylor and Xu, 2010; Zhao et al., 2012).
Nonetheless, there is little evidence on the impact of REM activities on future operating
performance around firms that meet/beat zero earnings and last year’s earnings. However,
13
ACCEPTED MANUSCRIPT
from the above empirical literature, since all earnings manipulation activities lead to higher
income in the current period, there are some consequences of these activities; these could
either be opportunistic choices or signalling choices of managers, and thus, the results are
mixed. However, if firms tend to manage earnings for opportunistic reasons and manipulate
deteriorate in the future. On the other hand, manipulations of operating activities are less
PT
likely to significantly affect the operations of firms that are in strong financial and market
RI
positions and that intend to use earnings to communicate favourable private information
SC
about future performance. Compared to other firms, firms that are just meeting/beating
important earnings benchmarks around zero earnings and last year’s earnings have higher
NU
incentives to engage in REM manipulation and are expected to experience negative (positive)
effects on their future performance. By taking three measures of REM for firms to meet zero
MA
discretionary expenses and production in the presence of meeting earnings benchmarks (i.e.,
PT
zero earnings and last year’s earnings) and future operating performance.
CE
discretionary expenses and production in the absence of meeting earnings benchmarks (i.e.,
AC
zero earnings and last year’s earnings) and future operating performance.
3.1. Sample
14
ACCEPTED MANUSCRIPT
Financial and market data were extracted from Datastream and Worldscope databases
for all firms listed on the London Stock Exchange (LSE).7 Data on the industry classification
were initially based on international standard industrial classification (ISIC). We start our
sample period in 2009, taking into account the possibility that managers may have been
incentivised to manipulate earnings in response to the 2008 financial crisis. The sample
ended in 2015 because of the requirement for data for two subsequent years’ performance.
PT
Following prior research, we exclude firms that operate in regulated industries (with SIC
RI
codes between 4400 and 5000) and financial institutions (with SIC codes between 6000 and
SC
6500). Since the anticipation models to estimate abnormal REM were realised by the 2-digit
SIC industry, we exclude industries with fewer than seven firms from the sample (Peasnell et
NU
al., 2005). After excluding firm-year observations without sufficient data to calculate all
measures of REM activity, control variables, and missing data, we ultimately had 4,487
MA
Table 1 shows the industry distribution of the final sample by presenting the number of
PT
observations and percentage for each industry (division), which consists of 930 firms from 49
industries. In addition, the sample firms come from a variety of industries, and we focus on
CE
several divisions. The most heavily represented industries are in manufacturing with 1,452
AC
firms (32.36%, 2-digit SIC code 20-39), and 32.74% are in services with 1,469 firms (2-digit
SIC code 70-89). This is followed by the mining division (21.68%, SIC code 10-14), the
retail trade division (7.24%, 2-digit SIC code 52-59), the division of construction (3.48%, 2-
digit SIC code 14-17), and the division of wholesale trade (2.5%, 2-digit SIC code 50-51).
7
To increase the sample size, this study does not restrict the sample to the firms with December fiscal year-end.
15
ACCEPTED MANUSCRIPT
Because we implement the standard models used to measure REM, we describe our
developed by Dechow et al. (1998) and employed by Roychowdhury (2006), Cohen et al.
(2008), Cohen and Zarowin (2010), Badertscher (2011) and Zang (2012) to construct our
earnings management metrics. Specifically, our REM measures are (1) abnormal levels of
cash flow from operations (Ab_CFO), (2) abnormal discretionary expenses (Ab_DISEX),
PT
and (3) abnormal production costs (Ab_PROD). We further aggregate these individual
RI
measures by computing two robust metrics of abnormal real activities to assess the overall
level of real activity manipulation. For the first aggregate measure, REM_1 — consistent
SC
with Cohen and Zarowin (2010) and Zang (2012) — we multiply Ab_DISEX by negative
NU
one and add it to Ab_PROD. A higher amount of this aggregate measure implies that suspect
increase reported earnings. For the second measure, REM_2 — again, consistent with Cohen
and Zarowin (2010) and Zang (2012) — Ab_CFO and Ab_DISEX are multiplied by negative
ED
one and then aggregated into one measure. For REM_1, we multiply it by negative one so
that the higher these amounts, are the more likely it is that the firm is engaging in sales-based
PT
study tests whether the two conflict effects of REM activities (“value destroying” and
“signalling”) of firms that just meet earnings benchmarks have an impact on subsequent
future performance. We estimate the following regression model with pooled ordinary least
16
ACCEPTED MANUSCRIPT
squares regressions and corrected the time-series cross-sectional dependencies in the data by
Most academic studies attempt to identify earnings management but do not provide
evidence on its magnitude; the current study addresses this by examining the relationship
PT
between the magnitude of REM proxies and future performance. In this model, the dependent
RI
variable is one-year-ahead industry-adjusted financial performance return on assets
SC
(Adj_ROAt+1) that is augmented with each REM activities’ measures, calculated as the
differences between firm-specific ROA and median ROA for the same year and industry (2-
NU
digit SIC code) as a direct measure of the firm’s longer-term cash flow.9 AREM refers to one
of the five measures for abnormal REM activities: the dependent variables Ab_CFO,
MA
Ab_PROD, Ab_DISEX and aggregate measures of real earnings management; REM_1 and
REM_2. The coefficients on AREM capture the main effects of abnormal REM activities on
ED
future performance. “Suspect” is defined as an indicator variable for suspect firm-years just
PT
meeting earnings benchmarks (i.e., zero benchmarks and last year’s benchmarks).
Suspect*AREM is an interaction term that captures the incremental effects of abnormal REM
CE
activities in the presence of meeting earnings benchmarks. Finally, to capture the total “value
destroying” and “signalling” impacts of REM activities, we test the sum of the coefficient on
AC
AREM, (??7) and the coefficient on the interaction terms of firms in the presence of just
8
The procedures of Fama-Macbeth are used as follows: in the first step, a time-series standard error regression
for each cross-sectional distribution of coefficients (e.g., firm- or portfolio-specific) is estimated. Then, in the
second step, the final coefficients’ estimates are obtained by basing inferences on the mean and standard
deviation of the resulting coefficients: in other words, the Fama-MacBeth t-statistics are based on the mean and
standard error of the time-series of coefficients from cross-sectional regressions.
9
In sensitivity analyses, we also examine two- and three-year-ahead performance and check the results with an
alternative firm’s performance variable (industry-adjusted return on equity).
17
ACCEPTED MANUSCRIPT
To avoid the problem of correlated omitted variables, we base our main set of control
variables on prior studies that suggest several factors that affect future operating performance
(e.g., Gunny, 2010; Taylor and Xu, 2010; Zhao et al., 2012; Leggett et al., 2016). To be
consistent with the dependent variable, all continuous independent variables are industry-
adjusted.10 The potential influence of a size effect is controlled by adding firm size (LnMVE)
PT
to the regression model, calculated as the natural logarithm of market capitalisation. Fama
RI
and French (1992) demonstrate that market capitalisation offers an important representation
SC
of the future expectations of the firm. To control for the life cycle of the firm, growth
opportunities are included in the regression models as a control variable because Fama and
NU
French (1992) note that growth opportunities are a direct signal of the relative future
prospects of firms and are calculated as the ratio of the market value of common equity
MA
divided by the book value of equity or market-to-book (MTB) ratio. Following Zang (2012),
the current study uses the bankruptcy prediction model developed by Altman (1968, 2000),
ED
which is represented here by the Altman Z_score, to control for the financial health of the
firm. A higher Z_score implies a firm’s healthier financial condition, and a lower Z_score
PT
Loss is measured as the indicator variable equal to one when net income before
extraordinary items is negative and zero otherwise and is included in the model because
AC
earnings are less persistent for firms with negative income.12 We also include the current-
period industry-adjusted financial profitability to control for the time-series properties (i.e.,
10
The variables are defined in Appendix B.
11
The Altman Z_score model is Z_score = 0.3*(X1) + 1.0*(X2) + 1.4*(X3) + 1.2*(X4) + 0.6*(X5), where
Z_score represents the Altman (1968, 2000) distress score with which the relative financial condition of the firm
can be explained based on magnitude and sign, measured at the beginning of year t. X1 represents the net
income before extraordinary items are added to the total assets ratio. X2 represents the total sales to the total
assets ratio. X3 represents the retained earnings to the total assets ratio. X4 represents the working capital to the
total assets ratio. X5 represents the market value of equity to the total liabilities ratio.
12
Roychowdhury (2006) finds evidence that firms with higher net incomes are less likely to manage earnings
and engage in REM.
18
ACCEPTED MANUSCRIPT
(Sloan, 1996; Dechow et al., 2003; Kraft et al., 2007, Gunny, 2010; Taylor and Xu, 2010).
Following Kothari and Sloan (1992) and Gunny (2010), to control for the association
between stock performance and future net income, market-adjusted abnormal returns is a
proxy for the firm’s market performance. This is included in the regression models as a
PT
control variable, calculated as the difference between monthly buy-and-hold raw returns and
RI
the monthly market buy-and-hold return, compounded over 12 months of the fiscal year t.13
SC
4. Empirical results
empirical models comparing suspect firm-years just meeting zero earnings and last year’s
MA
earnings with the non-suspect firm-years. The mean industry-adjusted return on assets (ROA)
are -10.3% and -10.4% for the subsequent dependent variable (Adj_ROAt+1) and current year
ED
industry-adjusted return on assets (Adj_ROA), respectively. The medians for both are, on
PT
average, zero, which is estimated because the variables are calculated by deducting the
The suspect firm-years in Panel A and Panel B have significantly lower values for
AC
Ab_DISEX than non-suspect firm-years, which suggests that suspect firm-years around zero
earnings and last year’s earnings with low discretionary expenses engage more in real activity
manipulation. In contrast, in Panel A, the mean REM_1 and REM_2 of suspect firm-years
have significantly higher means than non-suspect firm-years, suggesting that the firm also
engages in real activity manipulation. The mean Ab_CFO and Ab_PROD of suspect firm-
13
Consistent with Bens et al. (2002), this examination uses the 12 months buy-and-hold stock return on the
firm’s ordinary shares. Gunny (2010, p. 877) computes size-adjusted abnormal returns as “the monthly buy and
hold raw return minus the monthly buy and hold return on a size matched decile portfolio of firms compounded
over 12 months of fiscal year t”.
19
ACCEPTED MANUSCRIPT
years around zero earnings have a higher mean, although this is not significant compared to
non-suspect firm-years. In addition, in Table 2 (Panel B), the mean of Ab_CFO, Ab_PROD,
REM_1 and REM_2 of the suspect firm-years have a significantly higher mean than the
means for the non-suspect firm-years. This suggests that suspect firm-years that just meet
The mean of Z_score is 7.109, with a median of 2.753, above the cut-off point (1.80:
PT
Altman, 1968, 2000) of being a healthy firm. These values are comparable to those in Zang
RI
(2012). The mean of Loss is 0.416 with a median of zero. Approximately 41.6% of the
SC
sample observations experienced losses, suggesting that firms might have the potential to
engage in real manipulation. Finally, consistent with prior literature (e.g., Roychowdhury,
NU
2006; Cohen et al., 2008; Gunny, 2010; Zhao et al., 2012; Leggett et al., 2016), suspect and
non-suspect firm-years around zero earnings and last year’s earnings are different in
MA
Table 3 provides information about the Pearson and the Spearman correlation
ED
significantly negatively related to three of the five REM measures except REM_1,
PT
Ab_DISEX, indicating that the main effects of abnormal REM activities on the firm in the
CE
absence of just meeting important earnings benchmarks is that they perform worse in the
future or in signalling future performance. However, this reduces the firm’s value which will
AC
harm the firm’s future performance (value destroying), which is comparable to the findings
of prior studies. Addressing the correlation coefficients above 0.60 specifically, the
positively correlated with firm size (Pearson 33%, 30%, respectively and Spearman 43%,
20
ACCEPTED MANUSCRIPT
46%, respectively). This supports the assumption that firms have better current and future
negative correlation with the firms that experience negative net income (Loss), although this
The analysis of the correlation among the REM activity proxies reveals that the
correlation between Ab_PROD and Ab_DISEX is significantly positive (Pearson 35.5% and
PT
Spearman 52%). This suggests that managers are using real activity manipulation, which
RI
leads to abnormally high production costs that occur simultaneously with reducing
SC
discretionary expenditures (Gunny, 2010; Zang, 2012). There is a significant negative
relationship between Ab_DISEX and Ab_CFO (Pearson -43%, Spearman -17%); this shows
NU
that reduction of discretionary expenses frees up more cash flow for the firm. This result is
consistent with prior research (e.g., Cohen et al., 2008; Cohen and Zarowin, 2010). In
MA
addition, the positive and significant correlation coefficient (34%, Pearson and 42%,
Spearman) between Ab_CFO and Ab_PROD shows that (a) UK firms could engage in
ED
different REM methods at the same time to trigger higher reported earnings, and (b) it has a
Furthermore, the higher correlation coefficients of abnormal REM and the aggregate
CE
measures of REM (REM_1, REM_2) are expected because these aggregate measures of REM
are the sum of two proxies, suggesting that firms engage in real activity manipulation, which
AC
is consistent with prior research (Zang, 2012). Finally, the mean variance inflation factors
(VIF) for the independent variables used in the regression analysis of subsequent operating
performance of firms just meeting zero earnings and last year’s earnings, for all five
measures of REM, are all less than 1.50, suggesting that multicollinearity is not a concern in
this study.
21
ACCEPTED MANUSCRIPT
Table 4 reports the mean coefficient estimates from the Fama-MacBeth regressions.
The t-statistics are calculated using standard errors corrected for autocorrelation using the
dependent variable (Adj_ROA t+1) over the subsequent one year using the proxies of the
REM methods. The coefficients’ estimate for Adj_ROA is significantly different from zero
and positive in each of the REM methods. This indicates that current-period industry-adjusted
PT
financial profitability is positively associated with future industry-adjusted ROA (p-value <
RI
0.01), which is consistent with the findings from prior empirical studies in the US (e.g.,
SC
Gunny, 2010; Leggett et al., 2016).
The signs and significance of the control variables are consistent with the results from
NU
prior studies (e.g., Gunny, 2010; Taylor and Xu, 2010; Zhao et al., 2012 and Leggett et al.,
2016) with only a few exceptions. The coefficients’ estimate on LnMVE is significantly
MA
different from zero and positive in each of the REM methods, indicating that suspect firm-
years to meet zero earnings and last year’s earnings have better future performance, which is
ED
consistent with the findings of Gunny (2010) and Leggett et al. (2016). The coefficients’
estimate on MTB is statistically significant at the 0.01 level and is negative in each of the
PT
REM methods. This indicates that growth firms underperform in the future, which is
CE
The coefficient estimates on Loss in each of the REM methods of the regression model
AC
are all significantly negatively associated with future performance (p-value < 0.01). This
indicates that firms that engage in REM activities in the absence of meeting earnings
other firm-years. Zang (2012) points out that future profitability will be more negatively
affected by REM activities when firms are in poor financial health, whereas coefficients on
the Z_score are not significant in each of the REM methods, except for the coefficient on
22
ACCEPTED MANUSCRIPT
aggregate measures of real earnings management (REM_2), which is positive and statistically
significant at 0.002 (t = 2.69) at the 1% level. This finding is in line with a prior empirical
study by Gunny (2010). On the other hand, no coefficients on return significantly provide
Gunny (2010, p. 860) notes that “When examining the relation between future
PT
determined such that there is a factor that affects RM and also affects firms’ future
RI
performance (e.g., RM firm-years being representative of poor performance), then this study
SC
suffers from a potential correlated omitted variable problem”. Therefore, we carefully
control for the endogenous relationships between REM and firm future performance with a
NU
focus on REM; this focus is conditional upon an earnings management incentive, which is to
guard against the effects of conflicting explanations and the possibility of the omission of
MA
correlated variables. However, Hypothesis 1 focuses on firms that manipulate their sales,
beyond the expanded focus on all firms that engage in REM. Hence, the coefficient of
interest ??9 represents the performance of Suspect firms that is augmented with REM
PT
The interaction term (Suspect*AREM) in the first three columns of Table 4 (Panels A
and B), which captures the incremental effects of abnormal REM activities in the presence of
AC
meeting zero earnings and last year’s earnings, is significantly positively associated with
future operating performance at the 0.01 significance level, which is consistent with
Gunny’s (2010) and Zhao et al.’s (2012) findings; this suggests that managers who engage in
REM to meet zero earnings and last year’s earnings through sales-based manipulation,
discretionary expenses, and overproduction have better subsequent performance than other
firm-years. They also convey a signal of superior future performance to the market. In the last
23
ACCEPTED MANUSCRIPT
two columns of Table 4, Panels A and B report the results from the regression model with
and Suspect_zero*REM_2, are positive and statistically significant at 0.211 (t = 2.86) and
0.540 (t = 4.30), respectively. Additionally, the average coefficients on the two aggregate
PT
significant at 0.148 (t = 3.10) and 0.343 (t = 4.99), respectively. Hence, these results confirm
RI
that to meet zero earnings and last year’s earnings, managers of suspect firm-years who
SC
simultaneously engage in discretionary expenses-based manipulation, production costs-based
Furthermore, the results of the AREM coefficients indicate that the main effects of
MA
abnormal REM activities are all negative and significant (p-value < 0.01). This is comparable
to the findings of prior studies (Roychowdhury, 2006; Cohen and Zarowin, 2010) and
ED
suggests the general value destroying of the shareholders’ effect of abnormal REM activities
in the absence of meeting earnings benchmarks. In other words, firms that do not meet zero
PT
earnings and firms that do not sustain recent performance but engage in REM activities
CE
perform worse in the future or in signalling future performance than other firm-years.
Regarding the joint signalling effect documented by Gunny in her 2010 study, she
AC
reports that firms that utilise REM to meet earnings benchmarks exhibit significantly better
future performance than other REM firms that miss the targets, jointly signalling to the
market that these firms perform better. As shown in Table 4 (Panel A), however, the results
of the sum of ??7 and ??9, which captures the combined “value-destroying” and “signalling”
impacts of REM activities, show that the term (??7 + ??9) is significantly positive for three of
five measures of REM — Ab_DISEX, Ab_PROD, and REM_2 (p-value < 0.1 or p-value <
24
ACCEPTED MANUSCRIPT
0.01). This indicates that just to meet zero earnings, firm-years with engagement in REM
activities have better subsequent operating performance (have significantly higher industry-
adjusted ROA) than other firm-years with abnormal REM activities. This result is consistent
with joint signalling; that is, engaging in REM activities in the presence of meeting the
important earnings benchmarks to signal superior future performance (e.g., Gunny, 2010;
PT
In Panel B of Table 4, the results of the sum of the coefficients Suspect_last*AREM,
RI
??9 and AREM show that ??7 is significantly positive for one of five measures of REM
SC
(Ab_PROD) analyses at the 0.05 significance level. This indicates that firm-years to just meet
last year’s earnings that engage in REM activities have better subsequent operating
NU
performance than other firm-years with abnormal REM activities. In addition, in the case of
Ab_DISEX, the sum of coefficients ??7 (-0.338) and ??9 (0.171) is -0.166. The Wald test t-
MA
values of ??7 and ??9 are significantly negative at the 0.05 significance level (t = -2.18),
indicating that firm-years that engage in REM to meet last year’s earnings perform worse in
ED
the future or in signalling future performance than other firm-years with abnormal REM
PT
activities.
4.3.1. The sensitivity of the results to the period of two-year and three year-ahead industry-
adjusted return on assets
AC
The results in Table 5 (Panel A and B) reveal that, in general, are robust when using
Adj_ROAt+2 as the future performance measure and are consistent with the results from
Adj_ROAt+1 in Table 4 (Panel A and Panel B). For example, four of the five interaction terms
of the suspect firm-years of all measures of REM remain the same, and all are significantly
positively associated with future operating performance in year two at the 0.01 significance
level. The Ab_CFO that interacted with suspect firm-years (Suspect_zero, Suspect_last) are
25
ACCEPTED MANUSCRIPT
still significantly positive, where the significance levels drop from 1% to 5% and 10%,
The main effects of all abnormal REM activities remain the same, and all are negative
and significant with future operating performance in year two at the 0.01 significance level.
In addition, the combined coefficients on the interaction terms of abnormal REM (??7, ??9),
PT
which represents the full impact of REM on a firm’s subsequent two-year operating
RI
performance, are also positive and significant at the 0.01 significance level in the abnormal
SC
production cost analysis. However, the combined coefficients are negative and statistically
significant at the 0.05 significance level in the abnormal discretionary expenses analysis.
NU
Furthermore, the other combined coefficients of ??7 and ??9 maintain the same sign but are
no longer statistically significant. Overall, the results generally remain unchanged, suggesting
MA
that our findings are robust to the subsequent operating performance in year two as well as in
5. Conclusions
PT
In this study, we focus on a sample of firms listed on the LSE to examine whether there
is an association between UK firms that manipulate their sales, discretionary expenses and
CE
production to just meet certain earnings benchmarks and future operating performance. Our
AC
regression results show that UK firms that manipulate their earnings to meet zero earnings
and last year’s earnings are all significantly positively associated with future operating
performance, which is consistent with Gunny’s (2010) and Zhao et al.’s (2012) findings and
suggests that abnormal REM is intended to meet zero earnings and last year’s earnings and
14
The untabulated results reveal that, in general, are robust when using Adj_ROA t+3 as the future performance
measure, consistent with the results from Adj_ROAt+1 Adj_ROAt+2. However, the results remain unchanged for
three-year-ahead, suggesting that our findings are also robust to the subsequent operating performance in year
three as well as in their performance in years one and two.
26
ACCEPTED MANUSCRIPT
transmit a signal of superior future performance to the market. Furthermore, the results also
show that UK firms that manipulate their earnings are all negatively associated with future
operating performance, which is comparable to the findings of prior studies (Gunny, 2010;
Zhao et al., 2012). These results suggest a general value destroying outcome of shareholders’
effect of abnormal REM activities in the absence of just meeting important benchmarks.
PT
limitations identified in this study. One limitation of this study is that we only investigated
RI
motivation for REM, which is to meet important earnings benchmarks, such as avoiding
SC
report losses and sustaining recent profit performance. In reality, many other factors could
limitation is that there is a variety of factors not investigated in this research that can
MA
influence REM in a number of ways: for example, audit quality, corporate governance or
disclosure policies are less likely to engage in earnings management. Another limitation with
ED
the present study is that it examines the manipulation of common business activities, such as
sales, discretionary expenditures and production, but does not investigate other measures of
PT
investment in R&D, abnormal gains on sales of fixed assets and accrual earnings
management. Finally, because this study focuses on UK data, due to data access limitations,
AC
we do not explore the factors mentioned above, which are considered important areas in the
27
ACCEPTED MANUSCRIPT
Appendices
The first measure for REM activities is the abnormal cash flows from operations
PT
(Ab_CFO). Following previous studies (e.g., Roychowdhury, 2006; Cohen et al., 2008;
Cohen and Zarowin, 2010; Badertscher, 2011), sales-based manipulations are expected to
RI
lead to decreased current-period operating cash flows. We express the normal cash flows
SC
from operations as a linear function of sales revenue and change in sales revenue in the
current period using the model developed by Dechow et al. (1998) as implemented in
NU
Roychowdhury (2006). To estimate this model, we run the following cross-sectional
regression for each industry and year for all firms listed on the London Stock Exchange
MA
(LSE):
ED
where 𝐶𝐹𝑂𝑖,𝑡 is cash flows from operations for firm 𝑖 in the year 𝑡, defined as cash flows
PT
from operations divided by lagged total assets; TAi,t-1 is the total assets at the beginning of
CE
period 𝑡 for firm 𝑖; 𝑆𝐴𝐿𝐸𝑆𝑖,𝑡 is the sales revenue during period 𝑡 for firm 𝑖; ∆𝑆𝐴𝐿𝐸𝑆𝑖,𝑡 is the
change in sales revenue from year 𝑡 − 1 to 𝑡; 𝑖 is the firm; and 𝜀𝑖,𝑡 is the error term. For every
AC
firm on the LSE, Ab_CFO is computed as actual cash flows from operations minus the
normal level of cash flows from operations predicted from equation (A-1).
Following previous studies (e.g., Roychowdhury, 2006; Cohen et al., 2008; Cohen and
Zarowin, 2010; Badertscher, 2011; Zang, 2012), we model discretionary expenses as a linear
function of lagged sales and then estimate the following model to derive the normal levels of
28
ACCEPTED MANUSCRIPT
discretionary expenses for all firms listed on the LSE cross-sectionally for each industry and
year,
𝐷𝐼𝑆𝐸𝑋𝑖,𝑡 1 𝑆𝐴𝐿𝐸𝑆𝑖,𝑡−1
= 𝛼0 + 𝛽1 ∗ (𝑇𝐴 ) + 𝛽2 ∗ ( ) + 𝜀𝑖,𝑡, (A-2)
𝑇𝐴𝑖,𝑡−1 𝑖,𝑡−1 𝑇𝐴𝑖,𝑡−1
where 𝐷𝐼𝑆𝐸𝑋𝑖,𝑡 is the discretionary expenses that are defined as the sum of R&D,
PT
advertising, and selling, general and administrative expenses in year 𝑡 for firm 𝑖; and
𝑆𝐴𝐿𝐸𝑆𝑖,𝑡−1 is the sales revenue at the beginning of year 𝑡 for firm 𝑖. For every firm on the
RI
LSE, the Ab_DISEX is computed as the difference between the actual discretionary expenses
SC
and the normal level of discretionary expenses.
The third measure of REM activities is abnormal production costs (Ab_PROD). Studies
NU
such as Roychowdhury (2006), Cohen et al. (2008), Gunny (2010), Cohen and Zarowin
MA
(2010), Badertscher (2011), and Zang (2012) define production costs as the sum of the cost of
goods sold and change in inventory during the year, and they express the normal level of
ED
(A-3)
AC
where 𝑃𝑅𝑂𝐷𝑖,𝑡 is the sum of the cost of goods sold in year 𝑡 for firm 𝑖 and the change in
inventory from 𝑡 − 1 to 𝑡; and ∆𝑆𝐴𝐿𝐸𝑆𝑖,𝑡−1 is the change in sales revenue at the beginning of
year 𝑡 for firm 𝑖. The Ab_PROD is computed as the difference between the actual values of
production costs and the normal levels predicted from equation (A-3). Equation (A-3) is
29
ACCEPTED MANUSCRIPT
Appendix B Variables’ definitions, measurement and sources.
Variable Definition/Measurement Data sources
TA = Total assets (WC02999). Worldscope
SALES = Sales revenue (WC01001). Worldscope
IBEI = Income before extraordinary items (WC01551). Worldscope
CFO = Cash flows from operations (WC04860). Worldscope
COGS = Cost of goods sold (WC01051). Worldscope
INV
PROD
=
=
Inventories (WC02101).
Production costs; the sum of cost of goods sold and change in inventory.
P T Worldscope
R&D
ADV and SG&A
DISEX
=
=
=
Research and development expenses (WC01201).
Advertising and selling, general and administrative expenses (WC01101).
Discretionary expenses; the sum of R&D, ADV and SG&A.
R I Worldscope
Worldscope
Ab_CFO
Ab_DISEX
Ab_PROD
=
=
=
C
Abnormal cash flows from operations, where Ab_CFO is measured by the estimated residual from the regression equation (A-1).
S
Abnormal discretionary expenses, where Ab_DISEX is measured by the estimated residual from the regression equation (A-2).
Abnormal production costs, where Ab_PROD is measured by the estimated residual from the regression equation (A-3).
REM_1
REM_2
=
=
engaging in real activity manipulation.
N U
The sum of Ab_DISEX*(-1) and Ab_PROD; the higher the values of this aggregate measure are, the more likely it is that the firm is
The sum of Ab_CFO*(-1) and Ab_DISEX*(-1); the higher the values of this aggregate measure are, the more likely it is that the firm
ROA
MTB
=
=
is engaging in real activity manipulation.
M A
Return on assets; income before extraordinary items scaled by lagged total assets.
Market-to-book; the ratio of market value of equity (WC08001) to book value of equity (WC03501), measured at the beginning of Worldscope
D
year t.
LnMVE = Logarithm of the market value of equity, measured at the beginning of year t.
Suspect
Suspect_zero
=
=
T E
An indicator variable for suspect firm-years just meeting/beating important earnings benchmarks.
An indicator variable that is set equal to one if net income before extraordinary items scaled by lagged total assets is between 0 and Worldscope
Suspect_last =
P
0.005 and is set equal to zero otherwise, based on Roychowdhury’s (2006) criteria to identify suspect firm-years.
An indicator variable that is set equal to one if the change in net income before extraordinary items from the last year is between 0 and
E
0.01 and is set equal to zero otherwise, based on Gunny’s (2010) criteria to identify suspect firm-years.
Loss
Return
=
=
C C
An indicator variable equal to one when the net income before extraordinary items is negative and zero otherwise.
Market-adjusted abnormal returns are a proxy for the firm’s market performance, calculated as the difference between monthly buy-
and-hold raw returns and the monthly market buy-and-hold returns, compounded over 12 months of the fiscal year t.
Datastream
Z_score
Adj_ROAt+1
=
=
A
Measures the financial strength at the beginning of year t, computed as 0.3*(IBEI/TA) + 1.0*(SALES/TA) + 1.4*(Retained earnings
(WC03495)/TA) + 1.2*(Working capital (WC03151)/TA) + 0.6*(Market value of equity/Total liabilities (WC03351)).
One-year-ahead industry-adjusted financial performance return on assets, calculated as the differences between firm-specific ROA and
median ROA for the same year and industry (2-digit SIC code) as a direct measure of the firm’s longer-term cash flow.
Worldscope
ROE = Return on equity is the ratio of net income before extraordinary items divided by common equity. (WC08301). Worldscope
Adj_ROEt+1 = One-year-ahead industry-adjusted financial performance return on equity, calculated as the differences between firm-specific ROE
and median ROE for the same year and industry (2-digit SIC code).
30
ACCEPTED MANUSCRIPT
References
Alhadab, M., Clacher, I., & Keasey, K. (2015). Real and accrual earnings management and IPO
failure risk. Accounting and Business Research, 45(1), 55-92.
Altman, E. (1968). Financial ratios, discriminate analysis and the prediction of corporate
bankruptcy. Journal of Finance, 23(4), 589-609.
Altman, E. (2000). Predicting financial distress of companies: Revisiting the Z-score and ZETA
models. Working paper, New York University.
Baber, W., Fairfield, P., & Haggard, J. (1991). The effect of concern about reported income on
discretionary spending decisions: The case of research and development. The Accounting
PT
Review, 66(4), 818-829.
Badertscher, B. (2011). Overvaluation and the choice of alternative earnings management
mechanisms. The Accounting Review, 86(5), 1491-1518.
RI
Ball, R., Kothari, P., & Ashok, R. (2000). The effect of international institutional factors on
properties of accounting earnings. Journal of Accounting and Economics, 29(1), 1-51.
SC
Barton, J., & Simko, P. (2002). The balance sheet as an earnings management constraint. The
Accounting Review, 77, 1-27.
Bartov, E., Givoly, D., & Hayn, C. (2002). The rewards to meeting or beating earnings
expectations. Journal of Accounting and Economics, 33(2), 173-204.
NU
Bens, D., Nagar, V., & Wong, M. (2002). Real investment implications of employee stock
option exercises. Journal of Accounting Research, 40(2), 359-393.
Bhojraj, S., & Libby, R. (2005). Capital market pressure, disclosure frequency-induced
MA
earnings/cash flow conflicts, and managerial myopia. The Accounting Review, 8(1), 1-20.
Bhojraj, S., Hribar, P., Picconi, M., & McInnis, J. (2009). Making sense of cents: An
examination of firms that marginally miss or beat analyst forecasts. The Journal of
Finance, 64(5), 2359-2386.
D
Brown, L., & Higgins, H. (2001). Managing earnings surprises in the US Versus 12 other
countries. Journal of Accounting and Public Policy, 20(4/5), 373-398.
TE
Brown, L., & Higgins, H. (2005). Managers’ forecast guidance of analysts: International
evidence. Journal of Accounting and Public Policy, 24(4), 280-299.
Burgstahler, D., & Dichev, I. (1997). Earnings management to avoid earnings decreases and
P
De Jong, A., Merten, G., van der Poel, M., & van Dijk, R. (2014). How does earnings
management influence investor’s perceptions of firm value? Survey evidence from
financial analysts. Review of Accounting Studies, 19(2), 606-627.
Dechow, P., & Dichev, I. (2002). The quality of accruals and earnings: The role of accrual
estimation errors. The Accounting Review, 77(1), 35-59.
Dechow, P., Ge, W., & Schrand, C. (2010). Understanding earnings quality: A review of the
proxies, their determinants and their consequences. Journal of Accounting and
Economics, 50(2/3), 344-401.
Dechow, P., Kothari, S., & Watts, R. (1998). The relation between earnings and cash flows.
Journal of Accounting and Economics, 25, 133-168.
Dechow, P., Richardson, S., & Tuna, I. (2003). Why are earnings kinky? An examination of the
earnings management explanation. Review of Accounting Studies, 8(2), 355-384.
31
ACCEPTED MANUSCRIPT
DeFond, M., & Park, C. (1997). Smoothing income in anticipation of future earnings. Journal of
Accounting and Economics, 23(2), 115-139.
Degeorge, F., Patel, J., & Zeckhauser, R. (1999). Earnings management to exceed thresholds.
The Journal of Business, 72(1), 1-33.
Demski, J. (1998). Performance measure manipulation. Contemporary Accounting Research,
15(3), 261-285.
Eldenburg, L., Gunny, K., & Soderstrom, N. (2011). Earnings management using real activities:
Evidence from nonprofit hospitals. The Accounting Review, 86(5), 1605-1630.
Ewert, R., & Wagenhofer, A. (2005). Economic effects of tightening accounting standards to
restrict earnings management. The Accounting Review, 80(4), 1101-1124.
Fama, E., & French, K. (1992). The cross-section of expected stock returns. The Journal of
PT
Finance, 47(2), 427-465.
Fama, E., & MacBeth, J. (1973). Risk, return, and equilibrium: Empirical tests. Journal of
Political Economy, 81(3), 607-636.
RI
Francis, B., Hasan, I., & Li, L. (2016a). Abnormal real operations, real earnings management,
and subsequent crashes in stock prices. Review of Quantitative Finance and Accounting,
SC
46(2), 217-260.
Francis, B., Hasan, I., & Li, L. (2016b). A cross-country study of legal-system strength and real
earnings management. Journal of Accounting and Public Policy, 35(5), 477-512.
Ge, W., & Kim, J-B. (2014). Real earnings management and the cost of new corporate bonds.
NU
Journal of Business Research, 67(4), 641-647.
Graham, J., Harvey, C., & Rajgopal, S. (2005). The economic implications of corporate financial
reporting. Journal of Accounting and Economics, 40, 3-73.
MA
Gunny, K. (2010). The relation between earnings management using real activities manipulation
and future performance: Evidence from meeting earnings benchmarks. Contemporary
Accounting Research, 27(3), 855-888.
Han, S., Kang, T., Salter, S., & Yoo, Y. (2010). A cross-country study on the effects of national
D
Harrell, F. (2015). Regression modeling strategies with applications to linear models, logistic
and ordinal regression, and survival analysis, (2nd ed.). New York.
Healy, P., & Palepu, K. (1993). The effect of firms’ financial disclosure strategies on stock
P
Kim, J-B., & Sohn, B. (2013). Real earnings management and cost of capital. Journal of
Accounting and Public Policy, 32(6), 518-543.
Koh, K., Matsumoto, D., & Rajgopal, S. (2008). Meeting or beating expectations in the post-
scandals world: Changes in stock market rewards and managerial actions. Contemporary
Accounting Research, 25(4), 1067-1098.
Kothari, S., Mizik, N., & Roychowdhury, S. (2016). Managing for the moment: The role of real
activity versus accruals earnings management in SEO valuation. The Accounting Review,
91(2), 559-586.
Kothari, S. (2001). Capital markets research in accounting. Journal of Accounting and
Economics, 31(1/3), 105-231.
Kothari, S., & Sloan, R. (1992). Information in prices about future earnings: Implications for
earnings response coefficients. Journal of Accounting and Economics, 15(2/3), 143-171.
32
ACCEPTED MANUSCRIPT
Kraft, A., Leone, A., & Wasley, C. (2007). Regression-based tests of the market pricing of
accounting numbers: The mishkin tests and ordinary least squares. Journal of Accounting
Research, 45(5), 1081-1114.
Leggett, D., Parsons, L., & Reitenga, A. (2016). Real earnings management and subsequent
operating performance. IUP Journal of Operations Management, 15(4), 7-32.
Leuz, C., Nanda, D., & Wysocki, P. (2003). Earnings management and investor protection: An
international comparison. Journal of Financial Economics, 69(3), 505-527.
Mizik, N. (2010). The theory and practice of myopic management. Journal of Marketing
Research, 47, 594-611.
Newey, W., & West, K. (1987). A simple, positive semi-definite, heteroscedasticity and
autocorrelation consistent covariance matrix. Econometrica, 55(3), 703-708.
PT
Peasnell, K., Pope, P., & Young, S. (2005). Board monitoring and earnings management: Do
outside directors influence abnormal accruals? Journal of Business Finance and
Accounting, 32(7/8), 1311-1346.
RI
Roychowdhury, S. (2006). Earnings management through real activities manipulation. Journal
of Accounting and Economics, 42, 335-370.
SC
Sankar, M., & Subramanyam, K. (2001). Reporting discretion and private information
communication through earnings. Journal of Accounting Research, 39(2), 365-386.
Sloan, R. (1996). Do stock prices fully reflect information in accruals and cash flows about
future earnings? The Accounting Review, 71(3), 289-315.
NU
Subramanyam, K. (1996). The pricing of discretionary accruals. Journal of Accounting and
Economics, 22(1/3), 249-281.
Tan, C., & Jamal, K. (2006). Effect of accounting discretion on ability of managers to smooth
MA
http://ssrn.com/abstract=1013228.
Zhao, Y., Chen, K., Zhang, Y., & Davis, M. (2012). Takeover protection and managerial
myopia: Evidence from real earnings management. Journal of Accounting and Public
AC
33
ACCEPTED MANUSCRIPT
Table 1
Final sample classified by industry (SIC) division.
Number of
SIC Number of Percentage Cum.
Division group industry
division group observations of sample (%)
division group
SIC 10 – SIC
Mining 4 973 21.68 21.68
14
SIC 15 – SIC
Construction 3 156 3.48 25.16
17
SIC 20 – SIC
PT
Manufacturing 20 1,452 32.36 57.52
39
Wholesale SIC 50 – SIC
2 112 2.50 60.02
trade 51
RI
SIC 52 – SIC
Retail trade 8 325 7.24 67.26
59
SC
SIC 70 – SIC
Services 12 1,469 32.74 100.00
89
Total 49 4,487 100.00%
NU
MA
D
P TE
CE
AC
34
ACCEPTED MANUSCRIPT
Table 2
Descriptive statistics of suspect firm-years versus non-suspect firm-years.
Panel A: Descriptive statistics by suspect_zero firm-years versus non-suspect_zero firm-years.
Adj_ROAt+1
N
4,410
Mean
-0.103
Median
0.000
SD
0.412
Min
-2.689
Max
0.443
N
71
Mean
-0.040
Median
-0.221
SD
0.129
Min
-0.490
Max
0.442
N
3,409
R IMean
-0.100
Median
0.000
SD
0.404
Min
-2.689
Max
0.443
Mean
0.060***
Adj_ROA
Ab_CFO
3,557
3,547
-0.104
0.000
0.000
0.026
0.422
0.243
-2.837
-2.007
0.483
1.983
73
73
-0.017
0.007
-0.030
-0.001
0.047
0.098
-0.122
U
-0.244 SC
0.151
0.405
3,484
3,474
-0.105
-0.0001
0.000
0.026
0.426
0.245
-2.837
-2.007
0.483
1.983
0.088***
0.007
A
0.230
N
-0.861 0.436 2,980 0.002 -0.037 0.364 -3.006 2.871 -0.109***
M
0.051 0.212 -0.430 0.694 2,506 -0.0007 -0.007 0.335 -2.771 3.317 0.035
REM_1
REM_2
2,173
3,027
-0.012
-0.001
0.018
0.022
0.553
0.340
-4.351
-2.858
3.625
2.464
39
55
E D
0.119
0.105
0.097
0.127
0.410
0.245
-0.772
-0.581
1.130
0.629
2,134
2,972
-0.015
-0.003
0.018
0.021
0.555
0.341
-4.351
-2.858
3.625
2.464
0.134**
0.108***
Suspect_zero
LnMVE
3,557
4,487
0.020
17.839
0.000
17.486
0.141
2.349
0.000
13.403
E
1.000
24.356 PT
73
73
1.000
17.463
1.000
17.334
0.000
2.049
1.000
14.131
1.000
24.123
3,484
3,484
0.000
17.913
0.000
17.554
0.000
2.361
0.000
13.403
0.000
25.356
1.000
-0.450*
C
4.936
C
-14.694 29.137 73 1.377 0.969 1.258 -0.441 5.350 3,484 2.613 1.609 4.943 -14.694 29.137 -1.236***
Return
Z_score
4,336
3,483
0.087
7.109
-0.014
2.753
A
0.640
20.349
-0.892
-38.838
3.429
139.354
73
71
-0.032
3.381
-0.065
2.058
0.394
5.123
-0.813
-2.390
1.533
30.631
3,425
3,412
0.027
7.186
-0.035
2.780
0.560
20.540
-0.892
-38.838
3.429
139.354
-0.059
-3.805***
Loss 4,487 0.416 0.000 0.493 0.000 1.000 73 0.000 0.000 0.000 0.000 0.000 3,484 0.410 0.000 0.492 0.000 1.000 -0.410***
35
ACCEPTED MANUSCRIPT
Variable N Mean Median SD Min Max N Mean Median SD Min Max N Mean Median SD Min Max Mean
Adj_ROAt+1 4,410 -0.103 0.000 0.412 -2.689 0.443 387 0.006 0.011 0.141 -1.544 0.442 3,093
T
-0.113
P
-0.0005 0.420 -2.689 0.443 0.119***
Adj_ROA 3,557 -0.104 0.000 0.422 -2.837 0.483 393 0.009 0.013 0.129 -1.235 0.283
I
3,164
R
-0.117 -0.001 0.443 -2.837 0.483 0.126***
C
Ab_CFO 3,547 0.000 0.026 0.243 -2.007 1.983 393 0.032 0.033 0.132 -1.084 0.449 3,154 -0.004 0.024 0.253 -2.007 1.983 0.036***
Ab_DISEX
Ab_PROD
3,035
2,559
0.000
0.000
-0.383
-0.007
0.362
0.333
-3.006
-2.771
2.871
3.317
324
285
-0.097
0.040
-0.108
0.022
0.267
0.268
-0.705
U
-0.774
S 1.443
0.944
2,711
2,274
0.011
-0.005
-0.032
-0.010
0.370
0.340
-3.006
-2.771
2.871
3.317
-0.108***
0.045***
REM_1 2,173 -0.012 0.018 0.553 -4.351 3.625 232 0.122 0.106
A N
0.488 -1.405 1.179 1,941 -0.028 0.011 0.558 -4.351 3.625 0.150***
E D
1.000 1.000 0.000 1.000 1.000 3,164 0.000 0.000 0.000 0.000 0.000 1.000
T
LnMVE 4,487 17.839 17.486 2.349 13.403 24.356 393 18.952 18.836 2.319 13.510 24.356 3,164 17.773 17.440 2.328 13.403 24.356 1.179***
MTB 4,487
4,336
2.552
0.087
1.574
-0.014
4.936
0.640
-14.694
-0.892
E P
29.137
3.429
393
389
2.460
0.092
1.730
0.023
4.030
0.501
-14.694
-0.892
29.137
3.429
3,164
3,109
2.604
0.017
1.578
-0.044
4.997
0.564
-14.694
-0.892
29.137
3.429
-0.144
0.075***
C
Return
Z_score
Loss
3,483
4,487
7.109
0.416
2.753
0.000
20.349
A
0.493
C
-38.838
0.000
139.354
1.000
388
393
5.989
0.155
2.955
0.000
15.199
0.362
-13.016
0.000
139.354
1.000
3,095
3,164
7.249
0.432
2.727
0.000
20.903
0.495
-38.838
0.000
139.354
1.000
-1.260
-0.277***
36
ACCEPTED MANUSCRIPT
years are the 393 firm-years with a reported change in net income before extraordinary items from the last year between 0% and 1% in Panel B. Adj_ROAt+1 = One-year-
ahead industry-adjusted financial performance return on assets, calculated as the differences between firm-specific ROA and median ROA for the same year and industry (2-
digit SIC code). Adj_ROA = The current period industry-adjusted financial performance. Ab_CFO = Abnormal cash flows from operations, where Ab_CFO is measured by
the estimated residual from the regression equation (A-1). Ab_DISEX = Abnormal discretionary expenses, where Ab_DISEX is measured by the estimated residual from the
regression equation (A-2). Ab_PROD = Abnormal production costs, where Ab_PROD is measured by the estimated residual from the regression equation (A-3). REM_1 =
The sum of Ab_DISEX*(-1) and Ab_PROD; the higher the values of this aggregate measure are, the more likely it is that the firm is engaging in real activity manipulation.
T
REM_2 = The sum of Ab_CFO*(-1) and Ab_DISEX*(-1); the higher the values of this aggregate measure are, the more likely it is that the firm is engaging in real activity
manipulation. LnMVE = Logarithm of the market value of equity, measured at the beginning of year t. MTB = Market-to-book; the ratio of market value of equity to book
P
I
value of equity, measured at the beginning of year t. Return = Market-adjusted abnormal returns is a proxy for the firm’s market performance, calculated as the difference
between monthly buy-and-hold raw returns and the monthly market buy-and-hold return, compounded over 12 months of the fiscal year t. Z_score = Measures the financial
R
strength at the beginning of year t, computed as 0.3*(Net income before extraordinary items/Total assets) + 1.0*(SALES/Total assets) + 1.4*(Retained earnings/Total assets)
S C
+ 1.2*(Working capital/Total assets) + 0.6*(Market value of equity/Total liabilities). Loss = An indicator variable equal to one when net income before extraordinary items is
negative and zero otherwise. Differences in means (unequal variances) are tested using t-tests. To limit the influence of outliers, all continuous variables are winsorised at the
top and bottom 1% of their distribution.
N U
M A
E D
P T
C E
A C
37
ACCEPTED MANUSCRIPT
Table 3
Pearson and Spearman correlations coefficients among all variables in the operating performance regression.
VARIABLE (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)
(1) Adj_ROAt+1 1.000 0.702*** 0.033*** 0.089*** -0.420*** 0.107*** -0.160*** -0.028 -0.149*** 0.434*** 0.132*** 0.189*** 0.120*** -0.498***
(2) Adj_ROA 0.659*** 1.000 -0.390*** 0.106*** -0.480*** 0.131*** -0.171*** -0.019 -0.153*** 0.462***
T
0.143***
P
0.248*** 0.135*** -0.703***
I
(3) Suspect_zero 0.021 0.029 1.000 0.005 0.028 0.045*** 0.019 0.031 0.051*** -0.024 -0.071*** -0.013 -0.046*** -0.118***
R
(4) Suspect_last 0.093*** 0.094*** 0.005 1.000 -0.402*** 0.131*** 0.057*** 0.098*** 0.087*** 0.158*** 0.024 0.081*** 0.031 -0.177***
SC
(5) Ab_CFO -0.354*** -0.529*** -0.004 -0.046*** 1.000 -0.171*** 0.420*** 0.139*** 0.370*** -0.204*** -0.138*** -0.166*** -0.510*** 0.351***
(6) Ab_DISEX 0.148*** 0.339*** 0.040*** 0.093*** -0.429*** 1.000 0.520*** 0.860*** 0.770*** 0.071*** -0.133*** 0.010 -0.112*** -0.187***
(7) Ab_PROD
(8) REM_1
-0.209***
-0.002
-0.214*
0.086*
0.015
0.032
0.043***
0.084***
0.338***
-0.055***
0.355***
0.836***
1.000
0.809***
N U
0.829***
1.000
0.736***
0.877***
-0.008
0.023
0.119***
-0.151***
-0.014
-0.020
-0.043***
-0.092***
0.033
-0.069***
(9) REM_2
(10) LnMVE
(11) MTB
-0.075***
0.327***
-0.101***
-0.007
0.300***
-0.078*
0.042***
-0.027
-0.035***
0.071***
0.156***
-0.009
0.275***
-0.123***
0.054***
0.750***
0.039***
-0.104*** M A
0.639***
-0.009
-0.075***
0.850***
0.020
-0.117***
1.000
-0.027
-0.071***
-0.044***
1.000
0.136***
-0.167***
0.352***
1.000
-0.083***
0.254***
0.243***
-0.112***
0.154***
0.306***
0.030
-0.427***
-0.101***
PT
(13) Z_score -0.003 0.004 0.026 -0.019 0.087*** -0.071*** 0.057*** 0.003 -0.014 0.016 0.150*** -0.049*** 1.000 -0.024
(14) Loss -0.385*** -0.478*** -0.118*** -0.177*** 0.297* -0.148*** 0.058*** -0.036 0.052*** -0.422* 0.013 -0.102*** 0.190*** 1.000
A C
*, **, and *** denote statistical significance of the coefficient at the 10%, 5%, and 1% levels, respectively.
Table 3 shows Pearson and Spearman correlation coefficients of all variables in the future operating performance regression for the full sample of firm-year observations
(4,487 firm-year observations). Suspect_zero firm-years are the 73 firm-years with reported income before extraordinary items between 0% and 0.5% of lagged total assets.
Suspect_last firm-years are the 393 firm-years with reported change in net income before extraordinary items from the last year between 0% and 1%. Adj_ROAt+1 = One-
year-ahead industry-adjusted financial performance return on assets, calculated as the differences between firm-specific ROA and median ROA for the same year and
industry (2-digit SIC code). Adj_ROA = The current period industry-adjusted financial performance. Ab_CFO = Abnormal cash flows from operations, where Ab_CFO is
measured by the estimated residual from the regression equation (A-1). Ab_DISEX = Abnormal discretionary expenses, where Ab_DISEX is measured by the estimated
38
ACCEPTED MANUSCRIPT
residual from the regression equation (A-2). Ab_PROD = Abnormal production costs, where Ab_PROD is measured by the estimated residual from the regression equation
(A-3). REM_1 = The sum of Ab_DISEX*(-1) and Ab_PROD; the higher the values of this aggregate measure are, the more likely it is that the firm is engaging in real
activity manipulation. REM_2 = The sum of Ab_CFO*(-1) and Ab_DISEX*(-1); the higher the values of this aggregate measure are, the more likely it is that the firm is
engaging in real activity manipulation. LnMVE = Logarithm of the market value of equity, measured at the beginning of year t. MTB = Market-to-book; the ratio of market
value of equity to book value of equity, measured at the beginning of the year t. Return = Market-adjusted abnormal returns is a proxy for the firm’s market performance,
calculated as the difference between monthly buy-and-hold raw returns and the monthly market buy-and-hold return, compounded over 12 months of the fiscal year t.
T
Z_score = Measures the financial strength at the beginning of year t, computed as: 0.3*(Net income before extraordinary items/Total assets) + 1.0*(SALES/Total assets) +
1.4*(Retained earnings/Total assets) + 1.2*(Working capital/Total assets) + 0.6*(Market value of equity/Total liabilities). Loss = An indicator variable equal to one when net
P
distribution.
R I
income before extraordinary items is negative and zero otherwise. To limit the influence of outliers, all continuous variables are winsorised at the top and bottom 1% of their
S C
N U
M A
E D
P T
C E
A C
39
ACCEPTED MANUSCRIPT
Table 4
Regression analysis of subsequent operating performance of firms just meeting earnings
benchmarks.
Panel A: Regression analysis of subsequent operating performance of firms just
meeting zero earnings using Roychowdhury’s (2006) criteria to identify the suspect
firm-years of real earnings management activities.
PT
-0.378*** -0.316*** -0.296*** -0.373*** -0.359***
Intercept
(-8.70) (-23.39) (-9.52) (-7.70) (-11.10)
RI
0.546*** 0.595*** 0.634*** 0.597*** 0.577***
Adj_ROA
(9.15) (17.05) (12.03) (12.32) (11.91)
SC
0.020*** 0.019*** 0.016*** 0.021*** 0.021***
LnMVE
(8.56) (17.12) (8.71) (8.15) (11.14)
-0.005*** -0.007*** -0.004*** -0.005*** -0.007***
MTB
NU
(-5.85) (-4.74) (-4.72) (-5.46) (-5.85)
0.002 -0.004 0.003 -0.002 -0.004
Return
(0.31) (-0.67) (0.40) (-0.34) (-0.75)
MA
Suspect_zero
(-2.06) (-5.17) (-6.68) (-2.46) (-6.52)
0.900*** 0.402*** 0.341*** 0.211*** 0.540***
Suspect_zero*AREM
CE
40
ACCEPTED MANUSCRIPT
PT
Adj_ROA
(9.12) (17.16) (12.06) (12.35) (11.97)
0.021*** 0.019*** 0.016*** 0.021*** 0.021***
LnMVE
RI
(8.53) (17.26) (8.84) (8.04) (10.75)
-0.005*** -0.007*** -0.004*** -0.006*** -0.007***
MTB
SC
(-5.90) (-4.65) (-4.66) (-5.47) (-5.81)
0.002 -0.003 0.003 -0.003 -0.004
Return
(0.31) (-0.63) (0.39) (-0.39) (-0.75)
NU
0.002 0.004 -0.003 0.002 0.002**
Z_score
(1.02) (0.24) (-1.07) (0.13) (2.57)
-0.027*** -0.046*** -0.029*** -0.033*** -0.034***
Loss
MA
Suspect_last
(0.37) (6.48) (4.48) (1.61) (2.83)
E
*, **, and *** denote statistical significance of the coefficient at the 10%, 5%, and 1% levels,
respectively.
Table 4 reports the estimation results of Fama-Macbeth regressions for the full sample of firm-year
observations (4,487 firm-year observations) explaining the dependent variable subsequent operating
performance (Adj_ROAt+1) of firms just meeting zero earnings and last year’s earnings using the proxies
of the REM methods. The dependent variable (Adj_ROAt+1) = One-year-ahead industry-adjusted financial
performance return on assets, calculated as the differences between firm-specific ROA and median ROA for the
same year and industry (2-digit SIC code). Adj_ROA = The current period industry-adjusted financial
performance. AREM = Refers to one of the five measures for abnormal REM activities; Ab_CFO =
Abnormal cash flows from operations, where Ab_CFO is measured by the estimated residual from the
regression equation (A-1); Ab_DISEX = Abnormal discretionary expenses, where Ab_DISEX is measured by
the estimated residual from the regression equation (A-2); Ab_PROD = Abnormal production costs, where
Ab_PROD is measured by the estimated residual from the regression equation (A-3); REM_1 = The sum of
41
ACCEPTED MANUSCRIPT
Ab_DISEX*(-1) and Ab_PROD; the higher the values of this aggregate measure are, the more likely it is that
the firm is engaging in real activity manipulation; REM_2 = The sum of Ab_CFO*(-1) and Ab_DISEX*(-1);
the higher the values of this aggregate measure are, the more likely it is that the firm is engaging in real activity
manipulation. Suspect = An indicator variable for suspect firm-years just meeting earnings benchmarks. Each
panel reports the estimation results using a different definition of suspect firm-years as discussed below. Panel A
reports the results for suspects just meeting the zero earnings (Suspect_zero), which are firm-years with earnings
before extraordinary items over lagged assets between 0 and 0.005. Panel B reports results for suspects just
meeting last year’s earnings (Suspect_last), which are firm-years with the change in net income from the last
year between 0 and 0.01. Suspect*AREM = An interaction term that captures the incremental effects of
abnormal REM activities relative to the presence of just meeting benchmarks. LnMVE = Logarithm of the
market value of equity, measured at the beginning of year t. MTB = Market-to-book; the ratio of market value
of equity to book value of equity, measured at the beginning of year t. Return = Market-adjusted abnormal
returns is a proxy for the firm’s market performance, calculated as the difference between monthly buy-and-hold
PT
raw returns and the monthly market buy-and-hold return, compounded over 12 months of the fiscal year t.
Z_score = Measures the financial strength at the beginning of year t, computed as: 0.3*(Net income before
extraordinary items/Total assets) + 1.0*(SALES/Total assets) + 1.4*(Retained earnings/Total assets) +
1.2*(Working capital/Total assets) + 0.6*(Market value of equity/Total liabilities). Loss = An indicator variable
RI
equal to one when net income before extraordinary items is negative and zero otherwise. To limit the influence
of outliers, all continuous variables are winsorised at the top and bottom 1% of their distribution. Note that t-
statistics are calculated using standard errors corrected for autocorrelation and possible heteroscedasticity
SC
using the Newey-West procedure. They are reported in parentheses. To be consistent with the dependent
variable, all continuous independent variables are industry-adjusted. To limit the influence of outliers, all
continuous variables are winsorised at the top and bottom 1% of their distribution.
NU
MA
E D
PT
CE
AC
42
ACCEPTED MANUSCRIPT
Table 5
Regression analysis of subsequent operating performance in year two of firms just meeting earnings
benchmarks.
Panel A: Industry-adjusted return on assets (Adj_ROA t+2) is a proxy for operating performance of firms
that just meet zero earnings.
PT
0.537*** 0.578*** 0.621*** 0.584*** 0.554***
Adj_ROA
(7.90) (10.72) (8.37) (10.24) (8.47)
0.021*** 0.020*** 0.017*** 0.021*** 0.022***
RI
LnMVE
(6.28) (7.98) (6.84) (9.39) (8.49)
-0.005*** -0.007*** -0.004*** -0.006*** -0.007***
MTB
SC
(-3.97) (-4.18) (-3.74) (-6.24) (-5.15)
0.001 -0.004 0.002 -0.002 -0.003
Return
(0.24) (-0.46) (0.19) (-0.23) (-0.34)
0.0002 0.0001 -0.0002 0.0005 0.0002
NU
Z_score
(0.89) (0.03) (-0.81) (0.21) (1.61)
-0.030*** -0.053*** -0.035*** -0.038*** -0.043***
Loss
(-2.46) (-3.89) (-2.56) (-3.06) (-5.00)
MA
Suspect_zero*AREM
(2.03) (4.60) (10.23) (2.58) (4.08)
E
43
ACCEPTED MANUSCRIPT
Panel B: Industry-adjusted return on assets (Adj_ROA t+2) is a proxy for operating performance of firms
that just meet last year’s earnings.
PT
0.021*** 0.020*** 0.017*** 0.021*** 0.021***
LnMVE
(6.24) (7.81) (6.82) (6.13) (8.34)
-0.005*** -0.007*** -0.004*** -0.006*** -0.007***
MTB
RI
(-3.97) (-4.13) (-3.72) (-3.70) (-5.11)
0.001 -0.004 0.002 -0.002 -0.003
Return
(0.13) (-0.43) (0.20) (-0.22) (-0.32)
SC
0.0002 0.0002 -0.0003 0.0001 0.0002
Z_score
(0.85) (0.01) (-0.82) (0.19) (1.56)
-0.029** -0.049*** -0.033** -0.036*** -0.040***
Loss
NU
(-2.35) (-3.65) (-2.37) (-3.20) (-4.50)
-0.211*** -0.368*** -0.106* -0.121** -0.274***
AREM
(-6.04) (-4.46) (-1.87) (-5.85) (-4.77)
0.006 0.027*** 0.011*** 0.014 0.021***
MA
Suspect_last
(0.41) (5.39) (2.64) (1.20) (2.59)
0.458* 0.273** 0.165*** 0.145*** 0.251***
Suspect_last*AREM
(1.96) (2.54) (5.30) (2.62) (3.99)
No. of observations 3,346 2,860 2,407 3,354 2,852
ED
*, **, and *** denote statistical significance of the coefficient at the 10%, 5%, and 1% levels,
respectively.
Table 5 reports the estimation results of Fama-Macbeth regressions for the full sample of firm-year
observations (4,487 firm-year observations) explaining the dependent variable subsequent operating
AC
performance (Adj_ROA t+2) of firms just meeting zero earnings and last year’s earnings using the proxies
of the REM methods. The dependent variable (Adj_ROA t+2) = Two-year-ahead industry-adjusted financial
performance return on assets, calculated as the differences between firm-specific ROA and median ROA for the
same year and industry (2-digit SIC code). Adj_ROA = The current period industry-adjusted financial
performance. AREM = Refers to one of the five measures for abnormal REM activities; Ab_CFO =
Abnormal cash flows from operations, where Ab_CFO is measured by the estimated residual from the
regression equation (A-1); Ab_DISEX = Abnormal discretionary expenses, where Ab_DISEX is measured by
the estimated residual from the regression equation (A-2); Ab_PROD = Abnormal production costs, where
Ab_PROD is measured by the estimated residual from the regression equation (A-3); REM_1 = The sum of
Ab_DISEX*(-1) and Ab_PROD; the higher the values of this aggregate measure are, the more likely it is that
the firm is engaging in real activity manipulation; REM_2 = The sum of Ab_CFO*(-1) and Ab_DISEX*(-1);
the higher the values of this aggregate measure are, the more likely it is that the firm is engaging in real activity
manipulation. Suspect = An indicator variable for suspect firm-years just meeting earnings benchmarks. Each
panel reports the estimation results using a different definition of suspect firm-years as discussed below. Panel
A reports results for suspects just meeting the zero earnings (Suspect_zero), which are firm-years with earnings
before extraordinary items over lagged assets between 0 and 0.005. Panel B reports results for suspects just
44
ACCEPTED MANUSCRIPT
meeting last year’s earnings (Suspect_last), which are firm-years with the change in net income from the last
year between 0 and 0.01. Suspect*AREM = An interaction term that captures the incremental effects of
abnormal REM activities relative to the presence of just meeting benchmarks. LnMVE = Logarithm of the
market value of equity, measured at the beginning of year t. MTB = Market-to-book; the ratio of the market
value of equity to book value of equity, measured at the beginning of year t. Return = Market-adjusted abnormal
returns is a proxy for the firm’s market performance, calculated as the difference between monthly buy-and-hold
raw returns and the monthly market buy-and-hold return, compounded over 12 months of the fiscal year t.
Z_score = Measures the financial strength at the beginning of year t, computed as: 0.3*(Net income before
extraordinary items/Total assets) + 1.0*(SALES/Total assets) + 1.4*(Retained earnings/Total assets) +
1.2*(Working capital/Total assets) + 0.6*(Market value of equity/Total liabilities). Loss = An indicator variable
equal to one when net income before extraordinary items is negative and zero otherwise. To limit the influence
of outliers, all continuous variables are winsorised at the top and bottom 1% of their distribution. Note that t-
statistics are calculated using standard errors corrected for autocorrelation and possible heteroscedasticity
PT
using the Newey-West procedure. They are reported in parentheses. To be consistent with the dependent
variable, all continuous independent variables are industry-adjusted. To limit the influence of outliers, all
continuous variables are winsorised at the top and bottom 1% of their distribution.
RI
SC
NU
MA
ED
PT
CE
AC
45
ACCEPTED MANUSCRIPT
Highlights:
2- Our results show that UK firms that manipulate their earnings to meet zero earnings
and last year’s earnings are all significantly positively associated with future
operating performance.
PT
3- Our results show that UK firms that manipulate their earnings are all negatively
associated with future operating performance in the absence of meeting earnings
benchmarks.
RI
SC
NU
MA
ED
PT
CE
AC
46