The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1475-7702.htm
The impact of earnings extremity
on information content
of cash flow
Wael Mostafa
Department of Accounting, College of Business Administration,
University of Bahrain, Sakhir, Bahrain, and
Rob Dixon
Impact
of earnings
extremity
81
Received 22 March 2011
Revised 30 May 2012
2 August 2012
Accepted 23 September 2012
Durham Business School, University of Durham, Durham, UK
Abstract
Purpose – In contrast to recent US studies, almost all prior UK studies have not supported the
incremental information content of cash flow beyond earnings. In addition, to date no UK study has
addressed the effect of earnings extremity on the incremental information content of cash flow and
earnings whilst controlling for the extremity of cash flow. Therefore, and in order to assess the
generality of recent US findings, the aim of this study is to examine the incremental information
content of cash flow from operations and earnings and the effect of extreme earnings on the
incremental information content of cash flow from operations in the UK firms.
Design/methodology/approach – Based on market-based accounting research, this study uses
statistical associations between accounting data (earnings and cash flow) and stock returns to
assess/measure the incremental information content (value relevance) of cash flow and earnings
and the effect of extreme earnings on the incremental information content of cash flow and earnings.
The paper follows the recent methodology in this area that employs the level and change of cash flow
and earnings as an estimation of their unexpected components and isolates the extreme cash flow and
earnings apart from the moderate ones.
Findings – The results show that both earnings and cash flow from operations have incremental
information content beyond each other. It is also found that extreme earnings lead to incremental
information content for only moderate (not extreme) cash flow. These results are consistent with the
findings of the recent US studies.
Practical implications – Overall, the findings of this study support the usefulness of using cash
flow information, in addition to earnings in firm valuation by investors in the UK market, especially
when earnings are extreme and cash flow is moderate. The accounting interpretation of these results,
in terms of disclosure of earnings components, is discussed.
Originality/value – The study makes the following contributions to the incremental information
content of cash flow and earnings literature in the UK. First, this study employs actual cash flow data
derived from cash flow statements. Second, none of the prior UK studies shares the current research
focus, which is to examine the effect of earnings extremity on the incremental information content
of cash flow and earnings whilst controlling for the extremity of cash flow itself. Third, this study
employs a large sample size for a more recent period.
Keywords Stock returns, Incremental information content, Moderate earnings, Extreme earnings,
Moderate cash flow, Extreme cash flow, United Kingdom, Investors
Paper type Research paper
1. Introduction
Cash flow from operations is an important indicator of a firm’s financial health.
The users of financial statements now have cash flow information after the statement
Review of Accounting and Finance
Vol. 12 No. 1, 2013
pp. 81-104
q Emerald Group Publishing Limited
1475-7702
DOI 10.1108/14757701311295845
RAF
12,1
82
of cash flow became part of the full set of financial statements. But questions remain as
to whether cash flow information is incrementally informative when compared to
accruals information for assessing a firm’s current as well as future performance.
Since 1986 the incremental information content of cash flow and accruals studies
have occupied an important place in accounting research, perhaps because of the original
arguments related to the importance of cash flow accounting versus accrual accounting.
This research evaluates the usefulness of disclosing cash flow data. More specifically,
studies on incremental information content of cash flow and accruals have been
conducted to investigate whether cash flow from operations and accruals information are
valued differently by investors. As a result, we can thus determine whether
disaggregated disclosures of cash flow and accruals information are more informative.
The literature has explored extensively the incremental information content of cash
flow and accruals. Overall, the empirical work in this area represents a concerted effort
to examine the role of cash flow information in firm valuation (Neill et al., 1991).
The earliest studies, which were published in the 1980s, on the incremental information
content of cash flow from operations and accruals provided only limited support or
mixed and inconclusive results regarding the incremental information content of cash
flow and for its role in security valuation (e.g. in the USA, Schaefer and Kennelley,
1986; Wilson, 1986, 1987; Rayburn, 1986; Bowen et al., 1987; Bernard and Stober, 1989;
in the UK, Board and Day, 1989; Board et al., 1989).
The focus of recent US work on the incremental information content of cash flow and
earnings has been distinguished by using a research methodology that incorporates
conceptual factors which may be relevant in clarifying the role of cash flow and earnings
in explaining security returns. This line of research was triggered by earnings return
relationship studies which suggested that the magnitude of the earnings response
coefficient may be affected by a number of factors such as earnings permanence
(Charitou et al., 2001). Based on this methodology, more recently however, researchers
have begun to explore the effect of earnings extremity on the incremental information
content of cash flow from operations and earnings, and develop a more accurate proxy
for the unexpected amount of cash flow and earnings. In these recent studies, the level
and change of cash flow and earnings were employed as an estimation of their
unexpected components. Furthermore, extreme cash flow from operations and earnings
were isolated apart from moderate ones based upon those extreme earnings and cash
flow have limited information content. This recent work on the incremental information
content of cash flow and earnings provides significant results regarding the incremental
information of cash flow beyond that contained in earnings, especially in the case of
studies that have examined the effect of earnings extremity on the incremental
information content of cash flow from operations and earnings (Ali, 1994; Cheng et al.,
1996; Cheng and Yang, 2003).
To date in the UK, almost no studies have documented incremental information
content for cash flow from operations similar to the recent US studies (Ali, 1994;
Cheng et al., 1996; Cheng and Yang, 2003). The following points can be noted regarding
all previous UK studies on the incremental information content of cash flow and
accruals in comparison with US research in this area. First, all of them have employed
an estimate for cash flow since reporting cash flow was not required before 1991
(FRS1). These estimates of cash flow were poor proxies for actual cash flow
(Bahnson et al., 1996; Cheng et al., 1997) and thus the issue of whether cash flow from
operations has incremental information content beyond accruals information is
unresolved yet in UK and the direct evaluation of FRS No. 1 disclosures is still
required. Second, no studies in the UK have addressed the effect of earnings extremity
on the incremental information content of cash flow and earnings whilst controlling for
the extremity of cash flow itself, equivalent to Cheng and Yang (2003) in the USA only
Charitou et al. (2001) in the UK examined the effect of earnings extremity on the
incremental information content of cash flow, but like Cheng et al. (1996) they did not
control for the extremity of cash flow itself. And in contrast to Cheng et al. (1996) and
Charitou et al. (2001) were not able to find incremental information content for cash
flow from operations when earnings are extreme. Third, the sample size of these UK
studies on the incremental information content of each flow and accruals was relatively
small. The sample size ranged from 39 British manufacturing firms over the period
from 1962 to 1977 in the study of Board and Day (1989) to 4,531 firm-year observations
for a sample of 197 British firms over a 23-year period covering the period from 1971 to
1993 in the study of Green (1999). Hence, these relatively small sample sizes limit the
generalisability of their conclusions.
These comments give rise to the objective of this study as an examination of
incremental information content of cash flow from operations and earnings and the
effect of extreme earnings on the incremental information content of cash flow and
earnings in UK firms. Based on the above comments, the principal differences between
this study and those of previous UK studies are as follows. First, this study uses
post-FRS No. 1 information derived from statements of cash flow. Using reported cash
flow information eliminates the need to develop a proxy for cash flow, and reduces the
measurement errors in the estimated figures of cash flow from operations employed in
prior UK research. Second, this study controls for the extremity of cash flow itself
when examining the effect of earnings extremity on the incremental information
content of cash flow and earnings. Without controlling for the extremity of cash flow,
the studies may fail to find incremental information content for cash flow (Cheng and
Yang, 2003). This is because extreme cash flow is less informative than moderate cash
flow (Ali, 1994). A final difference between this study and previous UK research in this
area is that the current study has employed a large simple size for a more recent period.
This large sample and the relatively recent period allow the possibility of generalizing
the results of this study.
Furthermore, the USA is more litigious and more rules based than the UK, which
would affect factors such as the timely recognition of losses and, hence, the effect of
extreme earnings. For example, Leuz et al. (2003) show that the variance of earnings
relative to cash flow is much higher for the USA than the UK. Therefore, it might be
interesting to determine if similar results to the USA, namely, that the market places
higher weight on cash flow when earnings are transitory than when earnings are
permanent, hold in a country with fewer extreme earnings events such as the UK.
In this study, when we examine the incremental information content of cash flow and
earnings, earnings are considered the primary profitability measure and cash flow from
operations is the secondary profitability measure. When earnings have less information
content, the market will look for another measure that has more information content and
a good surrogate measure for earnings is cash flow from operations (Cheng and Yang,
2003). However, when cash flow from operations is extreme and has less information
content, the market will depend upon the primary profitability measure, which is
Impact
of earnings
extremity
83
RAF
12,1
84
earnings, irrespective of whether earnings themselves are extreme or moderate. On this
basis, the current study investigates the effect of earnings extremity on the incremental
information content of cash flow from operations and earnings. As stated above, when
the effect of earnings extremity on the incremental information content of cash flow and
earnings is examined, we follow the same methodology of Cheng and Yang (2003) by
controlling for the extreme cash flow as well[1].
We hypothesize that both earnings and cash flow from operations have incremental
information content beyond each other. Regarding the effect of extreme earnings on the
incremental information content of cash flow from operations and earnings, the study
hypothesizes that when earnings are extreme the incremental information content of
cash flow from operations exists only for moderate cash flow and not for extreme cash
flow.
The results showed that first, without addressing the issue of the extremity, both
earnings and cash flow from operations have incremental information content beyond
each other. These findings are consistent with previous UK studies regarding the
incremental information content of earnings. However, in relation to the incremental
information content of cash flow from operations, and in contrast to all previous UK
studies in this area, this study is the first to provide strong evidence for the incremental
information content of cash flow from operations over earnings. Second, regarding the
effect of extreme earnings on the incremental information content of cash flow and
earnings, extreme earnings lead to incremental information content only for moderate
(not extreme) cash flow from operations. These results are consistent with the findings
of recent US studies (Cheng et al., 1996; Cheng and Yang, 2003).
These findings reveal that cash flow from operations and total accruals are valued
(associated with returns) differently from each other and moderate cash flow from
operations has higher valuation than extreme total accruals. Overall, these results
support the usefulness of both earnings and cash flow information in firm valuation by
investors in the UK market, especially, when earnings are extreme.
The next section reviews prior studies on the impact of earnings extremity on the
incremental information content of cash flow from operations and earnings and
develops research hypotheses. Section 3 discusses research method. Section 4 presents
variables definition and data selection. Section 5 provides empirical results, and
Section 6 concludes the paper.
2. Prior evidences and hypotheses developments
The relation between earnings and returns is affected by extreme earnings.
Specifically, extreme earnings have lower value relevance than moderate earnings
(Kormendi and Lipe, 1987; Collins and Kothari, 1989). Earnings may contain extreme
items with limited valuation implications. Examples of extreme items in earnings
include current and long-term accruals such as gains or losses on marketable
securities, or the foreign currency translation adjustment, losses due to restructuring,
current recognition (through asset sales) of previous periods (or the current period)
increases in market value, and one-time gains and losses from change in accounting
standards (Cheng et al., 1996; Christensen et al., 2005). Moreover, because compensation
contracts and debt covenants are often based on reported accounting earnings,
managers may attempt to introduce extreme gains and losses in earnings (Kothari,
2001). Cash flow also may contain extreme components. This is because managers may
intentionally defer or front-load the recognition of cash accompanying revenues or
expenses. Therefore, the smaller effect of extreme earnings on stock returns can also be
applied to extreme cash flow.
Freeman and Tse (1992) showed that earnings return relation is nonlinear and
transitory earnings have lower persistence than permanent earnings. They documented
a high-marginal price response after controlling for transitory components in earnings
financial analysts forecast errors. Scott (2003) argued that the reason for the expectation
of a higher earnings response coefficient (ERS) in the case of permanent earnings is that
the increase in the revenue or cost saving will persist, whereas the expectation of a lower
earnings response coefficient (ERS) in the case of transitory earnings exists because
there is no reason for these unexpected earnings to recur.
There are several ways for measuring the transitory components in earnings. Collins
and Kothari (1989), O’Hanlon et al. (1992), and Donnelly and Walker (1995) used the
time-series estimates to measure earnings persistence. Ali and Zarowin (1992b) and
Ou and Penman (1989) used earnings to price ratio to measure transitory components in
earnings. Freeman and Tse (1992) used the absolute change in earnings financial
analysts forecast errors deflated by the beginning of the period price to isolate transitory
components in earnings apart from permanent components. In general, Freeman and
Tse (1992) showed that the earnings response coefficient (ERC) is more sensitive to
forecast error magnitude than to firm-specific average persistence. They concluded that
measuring earnings permanence using the absolute change in earnings is better than the
time-series estimates because “investors may assign each earnings surprise a unique
persistence measure that depends on the absolute magnitude of the surprise” (p. 187).
Many studies have documented that moderate earnings have more information
content than extreme earnings (e.g. in the USA, Freeman and Tse, 1992; Ali and
Zarowin, 1992a, b; Das and Lev, 1994; in the UK, O’Hanlon et al., 1992; Donnelly and
Walker, 1995). Some other studies have found incremental information content for cash
flow from operations when earnings are extreme (Cheng et al., 1996; Cheng and Yang,
2003). In addition, other studies have reported that moderate cash flow has more
incremental information content than extreme cash flow (Ali, 1994).
In the USA, Ali (1994) extended prior work on earnings return relation, which
showed that extreme earnings have less information content than moderate earnings,
to incremental information content of cash flow and earnings studies. He tested the
incremental information content of cash flow from operations in the presence of its
extremity and his study showed that extreme cash flow is less informative than
moderate cash flow. In other words, Ali found that the market puts more weight on
moderate cash flow than on extreme cash flow. However, he did not control for the
effect of earnings extremity[2]. Cheng et al. (1996) examined the effect of earnings
extremity on the incremental information content of cash flow from operations and
earnings. Their study showed that the market places a higher weight on cash flow
when earnings are extreme than when earnings are moderate. However, they did
not control for the extremity of cash flow itself[3]. Pfeiffer et al. (1998) replicated
Ali’s (1994) study. They reported the same results of Ali (1994), either by using
component expectations from a random-walk model or from a model that incorporates
the historical auto- and cross-correlation structure in the components.
In the UK, Green (1999) investigated whether “the quality of earnings” as measured
by the firm-specific relationship between profitability and cash generating ability
Impact
of earnings
extremity
85
RAF
12,1
86
affects the securities market’s differential pricing of cash flow from operations and
current accruals. The results suggested that, when the firm-specific time-series
correlation between change in earnings and cash flow is low, the decomposition of
earnings into cash flow and accruals are value relevant[4]. Following Cheng et al. (1996)
study into the US market, Charitou et al. (2001) examined whether the incremental
information content of cash flow from operations increases when earnings are
transitory. Similar to Cheng et al. (1996), the results showed that the earnings response
coefficient diminishes when earnings are transitory. This means that there is a smaller
impact from transitory earnings on abnormal returns. In contrast to Cheng et al. (1996)
study, however, there is no corresponding increase in the cash flow coefficient.
Also in the USA, Cheng and Yang (2003) developed a methodology in which we can
isolate the moderate cash flow from operations and earnings apart from extreme ones.
Their methodology enabled them to measure the incremental information content of
cash flow and earnings when earnings are extreme, whilst controlling for the extremity
of cash flow, and vice versa. To measure the extremity of cash flow and earnings, they
combined the work of Ali (1994) and Cheng et al. (1996). This study supported the
incremental information content of moderate (not extreme) cash flow from operations
when earnings are extreme.
To sum up, earlier studies that did not address the effect of extreme earnings on the
incremental information content of cash flow have reported mixed and inconclusive
results regarding the incremental information content of cash flow from operations
(e.g. in the USA, Schaefer and Kennelley, 1986; Wilson, 1986, 1987; Rayburn, 1986;
Bowen et al., 1987; Bernard and Stober, 1989; in the UK, Board and Day, 1989;
Board et al., 1989). Recently however, at least three studies in the USA, Cheng et al.
(1996, 1997), and Cheng and Yang (2003), which employed actual cash flow figures,
showed that cash flow from operations showed a complementary role for earnings even
without addressing the issue of extreme earnings[5]. Based on the results of these
recent US studies and with the large and recent data and the use of actual numbers of
cash flow as shown in cash flow statement, this study addresses the two following
hypotheses in the UK market to provide conclusive results for the incremental
information content of cash flow beyond earnings similar to the recent US studies:
H1. Earnings have incremental information content beyond that contained in cash
flow from operations.
H2. Cash flow from operations has incremental information content beyond that
contained in earnings.
In respect of testing the effect of the earnings extremity on the incremental information
content of cash flow from operations and earnings, the results of recent US studies
showed that cash flow from operations may perform a role in explaining security
returns when the information content of earnings is limited by their transitory items.
In order to examine the effect of extreme earnings on the incremental information
content of cash flow, this study follows the work of Cheng and Yang (2003). Similar to
that work, the current study expects that when earnings are extreme the incremental
information content of cash flow will be conditioned on its extremity. In other words,
even if earnings are extreme, the incremental information content of cash flow does not
exist for the extreme cash flow from operations. It exists only for moderate cash flow
from operations and not for the extreme cash flow.
In addition, we would argue that given two signals (earnings and cash flow), we
would place more weight on the one with lower extremity. Therefore, if the earnings
signal is high extreme, we would weight cash flows more heavily. Based on that, the
extremity of earnings will affect the incremental information content of cash flow, and
in turn influence the strength of the relation between security returns and cash flow
information that will depend also on the persistence (transitory) of cash flow.
Based on the above discussion, the third hypothesis is as follows:
H3. When earnings are extreme the incremental information content of cash flow
from operations exists only for moderate cash flow and not for extreme cash
flow.
3. Research method
3.1 The incremental information content of cash flow and earnings (Models 1 and 2)
Recent research that examined the incremental information of cash flow and earnings
(Cheng et al., 1996; Charitou et al., 2001; Cheng and Yang, 2003) employs both change
and level of cash flow and earnings as proxies for their unexpected amounts. This is
because the level and change specification reduces the measurement error of the
unexpected amounts of cash flow and earnings when they have extreme components
(Ali and Zarowin, 1992a, b). Brown et al. (1987) and Ali and Zarowin (1992a) have
shown that if the level and the change variables are taken together as proxies for
measuring the unexpected amount of the variable, then the sum of the slope
coefficients on the change and level variables will be an estimate of the slope coefficient
of the unexpected amount of the variable. Hence, in the level and change combined
model, assessing the incremental information content of cash flow and earnings can be
performed by examining the statistical significance of the sum of the two slope
coefficients on the level and changes variables. According to Biddle et al. (1997), the
level and change combined model proposed by Easton and Harris (1991) is equivalent
to the one-lag version model[6].
Consistent with the recent research, this study employs the level and change
combined model (Model 2) to examine the incremental information content of earnings
and cash flow from operations. However, in order to provide a performance
benchmark for the level and change combined model, this study employs also the
change model (random-walk model) (Model 1). We operationalize Models 1 and 2 in the
following way.
Change model (Model 1):
Rit ¼ a0t þ a1t DE it þ a2t DCF it þ 1it ;
where Rit is the annual market adjusted stock returns of firm i measured over
the fifth month of year t to the fourth month of year t þ 1. DEit is the change in
earnings and DCFit is the change in cash flow from operations for firm i in year t.
Earnings and cash flow are deflated by the market value of equity at the beginning of
year t.
This study hypothesizes positive and significant values for a1t and a2t. A positive
and significant value of a1t and a2t means that earnings have incremental information
beyond that contained in cash flow from operations, and cash flow from operations has
incremental information beyond that contained in earnings, respectively.
Impact
of earnings
extremity
87
RAF
12,1
88
Level and change combined model (Model 2)[7]:
Rit ¼ a0t þ a1t DE it þ a2t DCF it þ a3t E it þ a4t CF it þ 1it ;
ð2Þ
where Eit is the level of earnings and CFit is the level of cash flow from operations for
firm i in year t. Change and level of earnings and cash flow are deflated by the market
value of equity at the beginning of year t. The sum (a1 þ a3) combines the estimated
coefficients of the change and level of earnings; (a2 þ a4) combines the estimated
coefficients of the change and level of the cash flow from operations.
This study hypothesizes positive and significant values for the sums (a1 þ a3) and
(a2 þ a4). A positive and significant value of (a1 þ a3) and (a2 þ a4) means that
earnings have incremental information beyond that contained in cash flow from
operations and cash flow from operations has incremental information beyond that
contained in earnings, respectively.
3.2 The effect of the extremity (Models 3 and 4)
Further analysis for the issue of the effect of extreme earnings on the incremental
information content of cash flow has focused on the level and change combined model.
We employ the contextual models with a dummy variable approach for measuring the
effect of the extremity. The contextual models (Models 3 and 4) are as follow:
Rit ¼ a0t þ a1t DE it þ a2t DCF it þ a3t E it þ a4t CF it þ a5t Dit £ DE it þ a6t Dit
£ DCF it þ a7t Dit £ E it þ a8t Dit £ CF it þ 1it
ð3Þ
where (a1 þ a3) and (a2 þ a4) are the sum of the estimated coefficients of the change
and level of earnings and cash flow from operations when earnings are moderate,
respectively. The sums (a5 þ a7) and (a6 þ a8) combine the estimated coefficients of
the change and level of earnings, and cash flow from operations conditioned on the
extremity of earnings (when earnings are extreme), respectively. D is an indicator
variable (0, 1) to determine moderate and extreme earnings, respectively.
As stated before, this study measures the effect of earnings extremity on the
incremental information content of cash flow whilst controlling for the extremity of
cash flow itself by using the approach of Cheng and Yang (2003) as follows.
To control for the extremity of cash flow from operations, we perform the analysis
by partitioning the data set into two sub-samples. The first sub-sample includes
moderate cash flow from operations, while the second includes extreme cash flow from
operations. The ratios of cash flow from operations to market value of equity at the end
of year t are used to determine moderate and extreme cash flow from operations.
In each year, we rank firms into nine groups by their ending-of-year cash flow from
operations-market value of equity ratios (CFit/MVit) with an approximately equal
number of firms per group, where the tenth group includes firms with only negative
cash flow from operations. We classify firms in the middle six groups (groups 3-8) as
moderate cash flow from operations (the first sub-sample) and firms in the remaining
four groups as extreme cash flow from operations (the second sub-sample).
Then in order to measure the effect of earnings extremity on the incremental
information content of cash flow and earnings for each sub-sample, we classify the
data set for each sub-sample (moderate cash flow sub-sample and extreme cash flow
sub-sample) into two sub-groups. The first sub-group includes moderate earnings,
while the second includes extreme earnings. The ratios of earnings to market value of
equity at the end of year t are used to determine moderate and extreme earnings.
In each year and for each sub-sample, we rank firms into nine groups by their
ending-of-year earnings-market value of equity ratios (Eit/MVit) with an approximately
equal number of firms per group where the tenth group includes firms with only
negative earnings. We classify firms in the middle six groups (groups 3-8) as moderate
earnings and firms in the remaining four groups as extreme earnings. In equation (3),
Dit ¼ 0 for firms with moderate earnings, and equals 1 for firms with extreme earnings.
When equation (3) is estimated separately for the moderate cash flow sub-sample,
we call it Model 3, and Model 4 when estimated separately for the extreme cash flow
sub-sample. Therefore, Model 3 is employed to examine the effect of extreme earnings
on the incremental information content of cash flow from operations for moderate cash
flow from operations sub-sample, while Model 4 is employed to examine the effect of
extreme earnings on the incremental information content of cash flow from operations
for extreme cash flow from operations sub-sample.
In Model 3, this study hypothesizes a positive and significant value for the sum
(a6 þ a8) and a negative and significant value for the sum (a5 þ a7). A positive and
significant value for (a6 þ a8) implies that, when earnings are extreme the incremental
information content exists for moderate cash flow from operations. A negative and
significant value for (a5 þ a7) implies that the extremity of earnings leads to a
negative effect on the incremental information content of earnings.
In Model 4, this study hypothesizes a negative and significant (or positive but not
significant) value for (a6 þ a8) and a negative and significant value for (a5 þ a7).
A negative and significant (or positive but not significant) value for (a6 þ a8) implies
that, even when earnings are extreme, the incremental information content does not
exist for extreme cash flow from operations. A negative and significant value for
(a5 þ a7) implies that the extremity of earnings leads to a negative effect on the
incremental information content of earnings.
For estimating the empirical models, this study follows the same methodology of
prior studies in this area. Change model (Model 1), level and change combined model
(Model 2), and the two contextual models with a dummy variable approach for
measuring the effect of the extremity (Models 3 and 4) are estimated in two ways:
cross-sectionally by year, and pooling the data cross-sectionally by year and
intertemporally. The first estimation method, cross-sectionally by year, is performed
by estimating all previous four models separately for each of the seven sample years,
1996-2002. There is potential for cross-sectional correlation problems in the residuals[8]
leading to similar consequences to those deriving from heteroskedasticity
(i.e. understating the standard errors and hence overstating t-statistics) (Green,
1999). Consequently, this study employs the mean test (cross-temporal t-statistics) as
suggested by Bernard (1987). In the mean test, the t-statistic is computed by dividing
the mean of the yearly coefficients by its standard error. Thus, the inferences are based
on the time series of the annual parameter estimates of the annual cross-sectional
regression[9]. The second estimation method, pooling the data cross-sectionally by
year and intertemporally, is performed by estimating all previous four models by
pooling the data for all seven sample years (1996-2002). To control for the potential
effects of heteroskedasticity and autocorrelation in the residuals, the pooled regression
is estimated under the white cross-section method, which is derived by treating the pool
Impact
of earnings
extremity
89
RAF
12,1
90
regression as a multivariate regression, with an equation for each cross-section,
and calculating white-type robust standard errors for the system of equations[10]
(EViews 5 User’s Guide, pp. 855-6).
4. Variables definition and data selection
Consistent with prior studies, the variables employed in this study are defined as
follows[11]. First, earnings are defined as net income before extraordinary items and
dividends (Worldscope item WC 01551)[12]. This item represents income before
extraordinary items and preferred and common dividends, but after operating and
non-operating income and expenses, reserves, income taxes, minority interest and
equity in earnings. Second, cash flow is defined as net cash flow from operating
activities (WorldScope item WC 04860). This item represents the net cash receipts and
disbursements resulting from the operations of the company. Third, the annual market
adjusted return is measured over the 12-month period beginning on the fifth month of
each fiscal year-end, assuming that UK listed firms have to release their financial
statements within four months from the fiscal year-end (a four month lag period). The
annual market adjusted return is defined as the difference between the annual raw
return[13] for firm i measured over the 12-month period beginning on the fifth month of
each fiscal year-end, and annual share return on Financial Times All Share Index
(FTALLSH), measured over the 12-month period beginning on the fifth month of each
fiscal year-end. Brown and Warner (1980) found that in many cases a simpler method
which does not explicitly adjust for marketwide factors or for risk, such as the market
adjusted return model, provides similar results in comparison with the market model
for estimating abnormal returns. Fourth, the market value of equity is defined as the
market closing price-year-end multiplied by the number of common shares
outstanding (Worldscope item WC 08001).
Financial data, accounting items and the return index (RI), of this study are obtained
from the DataStream database. The sample data consist of all British firms quoted on
the London Stock Exchange in addition to dead British firms[14] over the period from
1995 to 2002.
The following criteria are applied to determine the sample of this study:
.
Firms have accounting data or shares prices (RI) for at least one year over the
period of the study (1995-2002).
.
Firms in financial sectors are excluded.
.
Firms have not changed their financial year-end during the period 1995-2002[15].
The above criteria produce a sample of 1,916 firms in each year. Table I presents a
number of firms in this analysis.
An additional two criteria are imposed to determine the final sample of this study.
The first is the availability of the accounting items for determining explanatory
variables; the second is the availability of the monthly RI to calculate the dependent
variable. In addition, the two extreme percent of observations that lie above 99 percent
and below 1 percent of the distribution of changes in earnings or cash flow from
operations or annual market adjusted return are considered as outliers and are
excluded from the sample. After applying the previous two criteria and removing
outliers, the final sample for each year through the period of the study is shown in
Table II.
The data set comprises 6,851 firm-year observations for a sample of 1,634 British
firms over seven year periods from 1996 to 2002. It should be noted that cash flow data
are available on DataStream for a reasonable number of the British firms from 1995
upwards, but the study begins with the 1996 fiscal year. This is because the study
employs changes in accounting items as independent variables.
Impact
of earnings
extremity
5. Results
5.1 Descriptive statistics
In Panel A of Table III, the descriptive statistics indicate that the mean of earnings
(change in earnings) is less than the mean of cash flow from operations (change in cash
flow from operations). These results are expected because non-cash expenses such as
depreciation, and amortization reduces earnings and these items are added back to
cash flow from operations. The standard deviation of earnings (change in earnings) is
higher than that of cash flow from operations (change in cash flow from operations).
These results are not expected because earnings have the flexibility inherent in
generally accepted accounting principles that provide managers with the opportunity
to use accruals to smooth out the variations in cash flow across years. However, the
above values are fairly close.
From Panel B of Table III, the correlation matrix indicates that the market adjusted
return is more highly correlated with earnings than with cash flow from operations.
In addition, the market adjusted return is more highly correlated with the changes in
earnings than with changes in cash flow from operations. These results reveal that
earnings (changes in earnings) have a stronger association with the market adjusted
return than cash flow from operations (changes in cash flow from operations).
Furthermore, the values of the correlations coefficients between the explanatory
variables suggest that multicollinearity is not likely to be a problem for this study.
91
Number of firms
Initial sample
Exclude firms without data through entire period 1995-2002
Exclude financial firms
Exclude firms that changed their financial year-ends in the period 1995-2002
Sample size before excluding firms with insufficient data to calculate the study
variables
Year
1996
1997
1998
1999
2000
2001
2002
Totals
5,489
(3,039)
(263)
(271)
1,916
Sample size
Incomplete data
Outliers
Final sample size
1,916
1,916
1,916
1,916
1,916
1,916
1,916
13,412
948
841
808
879
964
874
856
6,170
51
61
58
54
55
52
60
391
917
1,014
1,050
983
897
990
1,000
6,851
Table I.
Sample size for firms
quoted on the
London Stock Exchange
and dead British firms
over the period 1995-2002
before excluding firms
that have insufficient
data to calculate the
study variables
Table II.
Final sample size of the
study for the sample of
1,634 British firms over
the seven-year period
1996-2002
RAF
12,1
92
Table III.
Descriptive statistics
on cash flow from
operations and earnings
Rit
DEit
Panel A: descriptive statistics of the variables
Mean
20.137
0.001
SD
0.513
0.184
Median
20.089
0.006
Minimum
22.89
21.147
Maximum
1.744
2.358
Panel B: correlation matrix among the variables
Rit
1.000
0.185 *
DEit
1.000
DCFit
Eit
CFit
DCFit
Eit
CFit
0.012
0.134
0.005
20.732
0.992
0.017
0.20
0.06
2 2.94
1.153
0.087
0.157
0.086
2 1.375
1.246
0.095 *
0.209 *
1.000
0.284 *
0.424 *
0.052 *
1.000
0.254 *
0.087 *
0.465 *
0.458 *
1.000
Notes: Statistical significance at: *0.01 level; the number of observations ¼ 6,851 firm-year
observations; Rit is the annual market adjusted stock return of firm i measured over the fifth month of
year t to the fourth month of year t þ 1; DEit (Eit) is the change (level) in earnings and DCFit (CFit) is the
change (level) in cash flow from operations for firm i in year t; all the explanatory variables are deflated
by the market value of equity at the beginning of year t; Pearson correlation coefficients are in the cells
above the diagonal
5.2 Empirical results[16]
5.2.1 Incremental information content of cash flow from operations and earnings
(Models 1 and 2). The results of the change model (Model 1) are reported in Table IV.
From Table IV, Panel A, the mean of the yearly coefficients of changes in earnings is
0.58 (t ¼ 7.98). The value 0.58 is significantly positive at the 1 percent level. The mean
of the yearly coefficients of changes in cash flow from operations is 0.25 (t ¼ 7.01).
The value 0.25 is significantly positive at the 1 percent level. These results indicate
incremental information content for earnings beyond cash flow from operations and for
cash flow from operations beyond earnings. The pooled-data analysis in Panel B
generates similar results.
The results of the level and change combined model (Model 2) are reported in
Table V. From Table V, Panel A, the mean of the summed yearly coefficients of the
level and change in earnings is 0.85 (t ¼ 9.31). The value 0.85 is significantly positive
at the 1 percent level. The mean of the summed yearly coefficients of the level
and change in cash flow is 0.46 (t ¼ 3.79). The value 0.46 is significantly positive at the
1 percent level. These results suggest that both earnings and cash flow from operations
have incremental information content beyond each other. The pooled-data analysis in
Panel B also reveals similar results that both earnings and cash flow from operations
possess incremental information content beyond each other.
Moreover, in the cross-sectional regression, the level and change combined model
(Model 2) has a larger mean adjusted R 2 (0.13) than does the change model (Model 1)
(0.05). The pooled-data analysis generates similar results. These results suggest that
including both the change and the level of cash flow and earnings in the regression
model is a better specification than the random-walk model.
The above findings are consistent with recent US studies that examined the
incremental information content of cash flow from operations and earnings (Cheng et al.,
1996; Cheng and Yang, 2003).
Year
n
Rit ¼ a0t þ a1t DE it þ a2t DCF it þ 1it
Coefficients (t-statistics)
a0
a1
Panel A: cross-sectional regressions
1996
917
2 0.09
0.62
1997
1,014
2 0.26
0.53
1998
1,050
2 0.23
0.93
1999
983
0.02
0.53
2000
897
2 0.07
0.47
2001
990
2 0.17
0.67
2002
1,000
2 0.16
0.31
Mean
2 0.14
0.58
t-value
( 2 3.73) * * *
(7.98) * * *
Panel B: pooled cross-sectional time-series regression
Pooled
6,851
2 0.14
0.48
t-value
( 2 4.08) * * *
(5.93) * * *
a2
Adj. R 2
0.2
0.25
0.33
0.16
0.24
0.4
0.14
0.25
(7.01) * * *
0.05
0.04
0.07
0.04
0.04
0.05
0.03
0.05
0.23
(5.07) * * *
0.04
Notes: Statistical significance at: *0.10, * *0.05, and * * *0.01 levels, respectively; n represents the
number of firm-year observations for each year, and for the total number of observation, respectively;
Rit is the annual market adjusted stock return of firm i measured over the fifth month of year t to the
fourth month of year t þ 1; DEit is the change in earnings and DCFit is the change in cash flow from
operations for firm i in year t; these variables are deflated by the market value of equity at the
beginning of year t; mean represents the mean of the seven-yearly coefficients, and the t-statistic of the
mean is obtained by dividing the mean by its standard error; significance levels at six degrees of
freedom are 3.143 (0.01 level), 1.943 (0.05 level) and 1.440 (0.10 level); in pooled cross-sectional timeseries regression, the White cross-section method is employed to control for the potential effects of
heteroskedasticity and autocorrelation in the errors
As a result, it is reasonable to conclude that the first two hypotheses, which stated that
earnings have incremental information content beyond that contained in cash flow
from operations and cash flow from operations has incremental information content
beyond that contained in earnings, are accepted. The accounting interpretation of these
results, in terms of disclosure of earnings components, reveals that cash flow from
operations and total accruals are valued (associated with returns) differently from each
other. In other words, these results indicate that cash flow from operations has a higher
valuation than total accruals. Hence, for a given amount of earnings, the stock market
responds more favorably to cash flow from operations than it does to total accruals.
Thus, investors prefer to observe both cash flow from operations and total accruals
separately.
5.2.2 Impact of earnings extremity on the incremental information content of cash
flow from operations and earnings, controlling for extreme cash flow (Models 3 and 4).
Table VI reports the results of the contextual model with a dummy variable approach
from the first sub-sample – moderate cash flow sub-sample (Model 3). In Panel A, the
mean of the summed yearly coefficients of the level and change in earnings in the
existence of its extremity (a5 þ a7) is 2 4.20 (t ¼ 2 12.05). The value 2 4.20 is
significantly negative at the 1 percent level. This result suggests a smaller impact from
extreme earnings on abnormal returns. The mean of the summed yearly coefficients of
the level and change in moderate cash flow from operations conditioned on the
extremity of earnings (interacted with earnings extremity) (a6 þ a8) is 1.89 (t ¼ 7.07).
Impact
of earnings
extremity
93
Table IV.
Incremental information
content of cash flow from
operations and earnings:
Model 1(change model):
change in cash flow from
operations and earnings
Table V.
Incremental information
content of cash flow from
operations and earnings:
Model 2 (level and change
combined model):
level and change of cash
flow from operations
and earnings
a0
0.56
0.95
0.68
20.05
0.49
0.74
0.42
0.54
(4.57 ) * * *
0.4
(8.79) * * *
0.18
2 0.04
0.08
0.23
0.13
2 0.57
2 0.05
2 0.01
( 2 0.06 )
2 0.08
( 2 0.83)
0.6
(2.56) * *
0.13
0.38
0.44
2 0.14
0.12
1.63
0.7
0.47
(2.13) * *
0.7
(11.31) * * *
0.85
1.16
1.06
0.52
0.74
1.01
0.6
0.85
(9.31) * * *
0.52
(3.6) * * *
0.31
0.34
0.52
0.09
0.25
1.06
0.65
0.46
(3.79) * * *
Sum of (a2 þ a4)
0.11
0.08
0.14
0.13
0.04
0.09
0.27
0.15
0.13
Adj. R 2
Notes: Statistical significance at: *0.10, * *0.05, and * * *0.01 levels, respectively; n represents the number of firm-year observations for each year and for
the total number of observation, respectively; Rit is the annual market adjusted stock return of firm i measured over the fifth month of year t to the fourth
month of year t þ 1; DEit (Eit) is the change (level) in earnings and DCFit (CFit) is the change (level) in cash flow from operations for firm i in year t; these
variables are deflated by the market value of equity at the beginning of yeart; mean represents the mean of the seven-yearly coefficients, and the t-statistic
of the mean is obtained by dividing the mean by its standard error; significance levels at six degrees of freedom are 3.143 (0.01 level), 1.943 (0.05 level), and
1.440 (0.10 level); (a1 þ a3) combines the estimated coefficients of the change and level of earnings; (a2 þ a4) combines the estimated coefficients of the
change and level of the cash flow from operations; in pooled cross-sectional time-series regression, the White cross-section method is employed to control
for the potential effects of heteroskedasticity and autocorrelation in the errors
Panel A: cross-sectional regressions
1996
917
20.14
0.29
1997
1,014
20.35
0.21
1998
1,050
20.29
0.38
1999
983
0.03
0.57
2000
897
20.08
0.25
2001
990
20.28
0.27
2002
1,000
20.18
0.18
Mean
20.19
0.31
t-value
( 2 3.66) * * *
(6.15) * * *
Panel B: pooled cross-sectional time-series regression
Pooled
6,851
20.2
0.3
t-value
( 2 4.19) * * *
(5.3) * * *
n
94
Year
Rit ¼ a0t þ a1t DE it þ a2t DCF it þ a3t E it þ a4t CF it þ 1it
Coefficients (t-statistics)
a1
a2
a3
a4
Sum of (a1 þ a3)
RAF
12,1
n
a0
a1
a2
a3
a4
a6
a7
Coefficients (t-statistics)
a5
a8
Sum of
(a1 þ a3)
Sum of
(a2 þ a4)
Sum of
(a5 þ a7)
Sum of
(a6 þ a8)
0.22
0.22
0.24
0.23
0.18
0.2
0.38
0.33
0.26
Adj.
R2
Notes: Statistical significance at: *0.10, * *0.05, and * * *0.01 levels, respectively; n represents the number of firm-year observations for each year and for the total number of observation,
respectively; Rit is the annual market adjusted stock return of firm i measured over the fifth month of year t to the fourth month of year t þ 1; DEit (Eit) is the change (level) in earnings and DCFit
(CFit) is the change (level) in cash flow from operations for firm i in year t; these variables are deflated by the market value of equity at the beginning of year t; mean represents the mean of the
seven-yearly coefficients, and the t-statistic of the mean is obtained by dividing the mean by its standard error; significance levels at six degrees of freedom are 3.143 (0.01 level), 1.943 (0.05 level)
and 1.440 (0.10 level); (a5 þ a7) combines the estimated coefficients of the change and level of earnings in the existence of its extremity; (a6 þ a8) combines the estimated coefficients of the
change and level of moderate cash flow from operations conditioned on the extremity of earnings; in pooled cross-sectional time-series regression, the White cross-section method is employed to
control for the potential effects of heteroskedasticity and autocorrelation in the errors; the whole sample of each year has been divided into two sub-samples: a moderate cash flow sub-sample
and an extreme cash flow sub-sample; the sample in Model 3 is the sub-sample of moderate cash flow from operations observations; the ratios of cash flow from operations to market value of
equity at the end of year t are used to determine moderate cash flow from operations; moderate cash flow from operations is defined by ranking all firms in each year into nine groups by their
ending-of-year cash flow from operations-market value of equity ratios with an approximately equal number of firms per group where the tenth group includes firms with only negative cash
flow from operations; the middle six groups are considered as moderate cash flow from operations; moderate cash flow from operations observations, in this sub-sample, are classified into two
groups; the first sub-group includes moderate earnings, while the second includes extreme earnings; the ratios of earnings to market value of equity at the end of year t are used to determine
moderate and extreme earnings; all firms in each year are ranked into nine groups by their ending-of-year earnings-market value of equity ratios with an approximately equal number of firms
per group where the tenth group includes firms with only negative earnings; the middle six groups are classified as moderate and the other four groups are classified as extreme; Dit ¼ 0 for
moderate firms and Dit ¼ 1 for extreme firms
Panel A: cross-sectional regressions
1996
559
20.4
0.67
20.28
4.75
20.18
2 0.74
0.36
24.36
2.57
5.42
20.46
25.1
2.93
1997
587
20.55
0.82
20.07
2.9
1.36
2 0.06
20.15
22.66
1.24
3.72
1.29
22.72
1.09
1998
569
20.49
0.61
20.11
3.47
0.96
2 0.36
0.4
23.19
1.36
4.08
0.85
23.55
1.76
1999
543
20.22
0.72
0.26
3.62
20.34
2 0.27
20.51
23.64
2.38
4.34
20.08
23.91
1.87
2000
466
20.28
0.75
0.11
4.53
20.03
2 0.35
20.32
24.18
1.6
5.28
0.08
24.53
1.28
2001
469
20.48
0.05
20.56
5.9
1.7
0.11
0.25
25.53
2.52
5.95
1.14
25.42
2.77
2002
476
20.31
20.28
0.31
5.13
0.39
0.4
20.51
24.57
2.05
4.85
0.7
24.17
1.54
Mean
20.39
0.48
20.05
4.33
0.55
2 0.18
20.07
24.02
1.96
4.81
0.5
24.2
1.89
( 2 0.42) (10.91) * * * (1.83) * ( 2 1.3) ( 2 0.45) ( 2 11.2) * * * (9.27) * * * (15.9) * * *
(2.01) * * ( 2 12.05) * * * (7.07) * * *
t-value
( 2 8.37) * * * (3) * *
Panel B: pooled cross-sectional time-series regression
Pooled 3,669
20.39
0.36
20.17
3.81
0.8
2 0.15
0.02
23.43
1.71
4.17
0.63
23.58
1.73
(0.1) ( 2 12.25) * * * (9.11) * * * (13.03) * * *
(2.08) * * ( 2 11.16) * * * (8.74) * * *
t-value
( 2 8.66) * * * (2.31) * * ( 2 1.1) (13.69) * * * (1.95) * ( 2 1.27)
Year
Rit ¼ a0t þ a1t DE it þ a2t DCF it þ a3t E it þ a4t CF it þ a5t Dit £ DE it þ a6t Dit £ DCF it þ a7t Dit £ E it þ a8t Dit £ CF it þ 1it
Impact
of earnings
extremity
95
Table VI.
Effect of extreme
earnings on the
incremental information
content of cash flow from
operations and earnings
whilst controlling for the
extremity of cash flow:
contextual model
with a dummy variable
approach regression
results for the sub-sample
of moderate cash flow
(Model 3)
RAF
12,1
96
The value 1.89 is significantly positive at the 1 percent level. This result suggests that
moderate cash flow from operations has a greater impact on abnormal returns in the
existence of extreme earnings. The pooled regression in Panel B generates similar
results.
Table VII reports the results of the contextual model with a dummy variable
approach from the second sub-sample – extreme cash flow sub-sample (Model 4).
The summed coefficients of the level and change in earnings in the presence of its
extremity (a5 þ a7) is significantly negative at the 1 percent level (2 2.51, t ¼ 2 4.28,
and 2 2.42, t ¼ 2 4.58) for the annual regressions and pooled regression in Panels A
and B, respectively. Similar to the results of Model 3 in Table VI, these results show a
significantly negative impact of extreme earnings on abnormal returns. The summed
coefficients of the level and change of extreme cash flow from operations conditioned
on the extremity of earnings (interacted with earnings extremity) (a6 þ a8) is
(0.27, t ¼ 1.64, and 0.30, t ¼ 2.02) for the annual regressions and pooled regression in
Panels A and B, respectively. The value 0.30 is significantly positive at the 5 percent
level and the value 0.27 is significantly positive at the 10 percent level. These results
indicate mixed and weak evidence of incremental information content for extreme cash
flow from operations when earnings are extreme.
Together, the results from Table VI (Model 3) and Table VII (Model 4) indicate that:
.
extreme earnings have negative and significant impact on abnormal returns
regardless of whether cash flow is moderate or extreme;
.
extreme earnings lead to incremental information content for moderate cash flow
from operations; and
.
extreme earnings do not lead to incremental information content for extreme
cash flow from operations[17].
These findings are consistent with the US study of Cheng and Yang (2003).
Following Clubb (2003) in his discussion of the findings of Cheng and Yang’s (2003)
study, we perform a relative value relevance analysis of cash flow component and total
accruals component of earnings[18]. Defining earnings as the sum of a cash flow from
operations and total accruals components, therefore:
.
the implied summed coefficients of the level and change of cash flow can be
calculated as the summed coefficients of the level and change of earnings plus
summed coefficients of the level and change of cash flow; and
.
the implied summed coefficients of the level and change of total accruals can be
calculated as the summed coefficients of the level and change of earnings.
From Table VI (Model 3), when earnings are moderate, the implied summed coefficients
of the level and change of cash flow is 5.31 (4.81 þ 0.50) and the implied summed
coefficients of the level and change of total accruals is 4.81. When earnings are extreme,
the implied summed coefficients of cash flow is 3.00 (4.81 þ 0.50 2 4.20 þ 1.89) and the
implied summed coefficients of total accruals is 0.61 (4.81 2 4.20). It can be seen that,
when earnings are extreme, there is a much sharper fall in the summed coefficient for
extreme total accruals component ((4.81 2 0.61)/4.81 ¼ 87 percent) than for moderate
cash flow component ((5.31 2 3)/5.31 ¼ 44 percent). These results mean that moderate
cash flow becomes more important than extreme total accruals when earnings
n
a0
a1
a2
2 0.41
0.11
2 0.23
2 0.71
0.07
0.66
0.37
2 0.02
( 2 0.11)
0.06
(0.36)
2.6
(4.63) * * *
a4
2.64
2.16
2.04
0.61
1.81
5.05
4.91
2.75
(4.41) * * *
a3
20.01
( 2 0.14)
0.17
20.46
20.41
0.09
20.3
0.001
0.37
20.08
( 2 0.64)
a5
20.07
( 2 0.5)
0.07
20.57
20.21
20.05
0.7
20.28
0.16
20.03
( 2 0.17)
a6
2 2.41
( 2 4.12) * * *
2 2.45
2 1.24
2 1.67
2 0.66
2 1.37
2 4.88
2 4.74
2 2.43
( 2 3.76) * * *
a7
0.37
(1.85) *
0.43
20.07
0.77
0.35
20.28
0.9
20.04
0.3
(1.75) *
a8
2.88
(5.35) * * *
2.88
2.63
2.8
0.94
2.11
5.43
4.73
3.08
(5.31) * * *
Sum of
(a1 þ a3)
0.12
(1.25)
20.12
0.48
20.14
20.15
20.15
0.34
0.36
0.09
(0.81)
Sum of
(a2 þ a4)
0.3
(2.02) * *
0.5
20.64
0.56
0.3
0.42
0.62
0.12
0.27
(1.64) *
2 2.28
2 1.7
2 2.08
2 0.57
2 1.67
2 4.88
2 4.37
2 2.51
( 2 4.28) * * *
2 2.42
( 2 4.58) * * *
Sum of
(a6 þ a8)
Sum of
(a5 þ a7)
0.11
0.12
0.2
0.13
0.04
0.09
0.28
0.13
0.14
Adj.
R2
Notes: Statistical significance at: *0.10, * *0.05, and * * *0.01, levels, respectively; n represents the number of firm-year observations for each year and for the total number of observation,
respectively; Rit is the annual market adjusted stock return of firm i measured over the fifth month of year t to the fourth month of year t þ 1; DEit (Eit) is the change (level) in earnings and DCFit (CFit)
is the change (level) in cash flow from operations for firm i in year t; these variables are deflated by the market value of equity at the beginning of year t; mean represents the mean of the seven-yearly
coefficients, and the t-statistic of the mean is obtained by dividing the mean by its standard error; significance levels at six degrees of freedom are 3.143 (0.01 level), 1.943 (0.05 level) and 1.440 (0.10
level); (a5 þ a7) combines the estimated coefficients of the change and level of earnings in the existence of its extremity; (a6 þ a8) combines the estimated coefficients of the change and level of
extreme cash flow from operations conditioned on the extremity of earnings; in pooled cross-sectional time-series regression, the White cross-section method is employed to control for the potential
effects of heteroskedasticity and autocorrelation in the errors; the whole sample of each year has been divided into two sub-samples: a moderate cash flow sub-sample and an extreme cash flow subsample; the sample in Model 4 is the sub-sample of extreme cash flow from operations observations; the ratios of cash flow from operations to market value of equity at the end of year t are used to
determine extreme cash flow from operations; extreme cash flow from operations is defined by ranking all firms in each year into nine groups by their ending-of-year cash flow from operationsmarket value of equity ratios with an approximately equal number of firms per group where the tenth group includes firms with only negative cash flow from operations; the first, second, ninth and
tenth groups are considered as extreme cash flow from operations; extreme cash flow from operations observations, in this sub-sample, are classified into two groups; the first sub-group includes
moderate earnings, while the second includes extreme earnings; the ratios of earnings to market value of equity at the end of year t are used to determine moderate and extreme earnings; all firms in
each year are ranked into nine groups by their ending-of-year earnings-market value of equity ratios with an approximately equal number of firms per group where the tenth group includes firms
with only negative earnings; the middle six groups are classified as moderate and the other four groups are classified as extreme; Dit ¼ 0 for moderate firms and Dit ¼ 1 for extreme firms
Panel A: cross-sectional regressions
1996
358
20.21
0.24
0.29
1997
427
20.43
0.47
0.37
1998
481
20.38
0.76
0.09
1999
440
0.03
0.33
0.56
2000
431
20.18
0.3
20.22
2001
521
20.47
0.38
20.32
2002
524
20.39
20.18
20.01
Mean
20.29
0.33
0.11
t-value
( 2 4.97) * * *
(3.09) * *
(0.9)
Panel B: pooled cross-sectional time-series regression
Pooled 3,182
20.31
0.28
0.06
t-value
( 2 4.85) * * *
(2.6) * * *
(0.52)
Year
Coefficients (t-statistics)
Rit ¼ a0t þ a1t DE it þ a2t DCF it þ a3t E it þ a4t CF it þ a5t Dit £ DE it þ a6t Dit £ DCF it þ a7t Dit £ E it þ a8t Dit £ CF it þ 1it
Impact
of earnings
extremity
97
Table VII.
Effect of extreme
earnings on the
incremental information
content of cash flow from
operations and earnings
whilst controlling for the
extremity of cash flow:
contextual model
with a dummy variable
approach regression
results for the sub-sample
of extreme cash flow
(Model 4)
RAF
12,1
98
are extreme. This confirms the above results of Model 3 regarding the existence of
incremental information content of moderate cash flow when earnings are extreme.
From Table VII (Model 4), when earnings are moderate, the implied summed
coefficients of the level and change of cash flow is 3.17 (3.08 þ 0.09) and the implied
summed coefficients of the level and change of total accruals is 3.08. When earnings are
extreme, the implied summed coefficients of the level and change of cash flow is 0.93
(3.08 þ 0.09 2 2.51 þ 0.27) and the implied summed coefficients of the level and
change of total accruals is 0.57 (3.08 2 2.51). These results indicate that there is not a
large difference in the fall of the summed coefficient of extreme total accruals
component ((3.08 2 0.57)/3.08 ¼ 82 percent) and extreme cash flow component
((3.17 2 0.93)/3.17 ¼ 71 percent) when earnings are extreme. Such a result means that
extreme cash flow and extreme total accrual are almost equivalently informative. This
confirms the above results of Model 4 regarding the lack of evidence of incremental
information content of extreme cash flow when earnings are extreme.
From all the above results, it can be concluded that the third hypothesis (which
stated that when earnings are extreme, the incremental information content of cash
flow from operation exists only for moderate cash flow and not for extreme cash flow)
is accepted. The accounting interpretation of these results, in terms of disclosure of
earnings components, reveals that moderate cash flow from operations and extreme
total accruals are valued (associated with returns) differently from each other. In other
words, these results indicate that moderate cash flow has a higher valuation than
extreme total accruals. Hence, for a given amount of extreme earnings, the stock
market responds more favorably and much more sharply to moderate cash flow from
operations than extreme total accruals. Thus, investors prefer to observe both
moderate cash flow from operations and extreme total accruals separately when
earnings are extreme. On the other hand, extreme cash flow from operations and
extreme total accruals are valued (associated with returns) equivalently. Hence, for a
given amount of extreme earnings, the stock market responds equivalently to extreme
cash flow from operations and extreme total accruals. Thus, investors do not prefer to
observe extreme cash flow from operations and extreme total accruals separately when
earnings are extreme. This is because the two components are being valued
equivalently and do not represent independent signals for the investor.
6. Concluding remarks
Based on UK data, this study investigates:
.
the incremental information content of cash flow from operations and earnings to
assess whether cash flow from operations and total accruals are valued
differentially; and
.
the effect of extreme earnings on the incremental information content of cash
flow from operations and earnings to assess whether extreme earnings lead to
incremental information content for cash flow from operations.
This study employs actual cash flow data derived from cash flow statements and a
large simple size, as well as control for the extremity of cash flow.
The results show that earnings have incremental information content beyond cash
flow from operations, and cash flow from operations has incremental information
content beyond earnings. Regarding the effect of extreme earnings on the incremental
information content of cash flow from operations and earnings, the results show that
extreme earnings lead to incremental information content for moderate (but not
extreme) cash flow from operations. These results are consistent with the findings of
the US study by Cheng and Yang (2003). Overall, these findings support the value
relevance of cash flow information, in addition to earnings by investors, to improve
the outcome of their decision regarding intelligent investment and credit decisions.
The research literature indicates:
.
a higher valuation of both current accruals and cash flow from operations than
non-current accruals[19] (this issue has been widely documented, e.g. Rayburn,
1986; Wilson, 1986, 1987; Pfeiffer et al., 1998; Pfeiffer and Elgers, 1999); and
.
weak evidence on the differential higher valuations of cash flow over current
accruals (e.g. in the USA, Rayburn, 1986; Bernard and Stober, 1989; Jennings,
1990; Ali, 1994; Pfeiffer et al., 1998[20]; and in the UK, Board and Day, 1989;
Board et al., 1989; Ali and Pope, 1994, 1995; Green, 1999).
Since this study has examined whether cash flow from operations, and total accruals,
as a whole, are associated with returns differently (which do not necessarily hold with
respect to assessing the differential valuations of cash flow from operations and
current accruals), a possible extension for this study is to extend the issue of the effect
of extreme earnings on cash flow to search for the effect of extreme working capital
from operations on the incremental information content of cash flow. This would
determine whether the role of moderate cash flow from operations in explaining the
value could be generalized when either earnings or working capital from operations
become extreme, and to assess whether a higher market valuation for moderate cash
flow from operations over extreme total accruals could be extended to the higher
valuation of moderate cash flow from operations over extreme current accruals.
Notes
1. This means that this study isolates extreme cash flow from operations apart from moderate
ones when examining the effect of extreme earnings on the incremental information content
of cash flow and earnings.
2. In other words, Ali (1994) did not examine the effect of extreme earnings on the incremental
information content of cash flow from operations.
3. In other words, Cheng et al. (1996) did not isolate extreme cash flow apart from moderate
ones when they examine the effect of extreme earnings on the incremental information
content of cash flow and earnings.
4. In this study “Earnings” was defined as net profit derived from normal trading activities
before depreciation and operating provisions, “cash flow” was defined as earnings minus
change in stock and change in total debtors and equivalents plus the change in creditors
during the year, and “current accrual” was defined as the difference between earnings and
cash flow. In this study, earnings and cash flow differ only with respect to the change in
working capital adjustment.
5. Recent work on incremental information content of cash flow and earnings (Cheng and
Yang, 2003) showed that it is easier for cash flow from operations to have incremental
information content beyond extreme earnings. The reason is that the market will value the
secondary measure, cash flow from operations, rather than the primary measure, earnings, if
this primary measure is extreme than the secondary measure.
Impact
of earnings
extremity
99
RAF
12,1
100
6. For more details about the equivalence between the level and change combined model and
the one-lag specification model, see Biddle et al. (1994, 1995, 1997).
7. Testing of significance for sum of two coefficients jointly has been performed by using a
Wald test of joint significance. t-statistics for sum of two coefficients (a1t þ a3t), for example,
is as follows. Computed T ¼ sum (a1t þ a3t)/the square root of {(variance of a1t þ variance
of a3t) 2 2 Covariance of (a1t, a3t)}. Where tabulated T is obtained by determined significant
level with (N-K) degrees of freedom: N ¼ the total number of observations and K ¼ the
number of right-hand side regressors in the equation of the regression. See Wald test in
Maddala (2001), Kennedy (2003), and Greene (2003). These statistical operations have been
performed directly by using EViews econometric software Version 5.
8. For more details about the potential problems of cross-sectional correlation in the residuals,
see Bernard (1987).
9. This procedure is known as Fama and Macbeth (1973) regression analysis. See also
Bernard’s (1987) paper on cross-sectional dependency.
10. For more details of white cross-sectional method, see: Wooldridge (2002) and Arellano (2003).
11. The definitions of these variables are according to the Worldscope items used to construct
the variables of this study.
12. (WC þ Number) is the code of Worldscope item. World Scope company accounts system has
been adopted by DataStream database since April 2003 onwards as a replacement of
DataStream company accounts data.
13. DataStream RI is used to calculate the returns instead of share price because it is adjusted for
dividends and capital actions such as share repurchases and share splits. Since The RI is
available on DataStream for each share and for the FTALLSH from 1988 onwards, this
variable is obtained directly from DataStream (DataStream code RI).
14. To avoid survivorship bias this study includes dead British firms in addition to surviving
firms. Dead firms are those which have merged, liquidated or become privately held.
15. Firms that have more than ten days’ difference between two consecutive year ends are
excluded. WC 05350 is the code of Worldscope item of firm’ fiscal year-end used to retrieve
the fiscal year-ends of firms.
16. For robustness checks, annual raw returns were employed as a dependent variable.
The empirical results are identical to those derived from using market adjusted returns as a
dependent variable.
17. For robustness checks to the results of Models 3 and 4, an alternative method for measuring
the extremity of cash flow from operations (earnings) was employed. This is based on whether
the absolute value of changes in cash flow (earnings) deflated by the market value of equity
at the beginning of year t lie above or below the yearly median. Firms falling below the median
are classified as moderate and firms falling above the median as extreme. The results have not
changed. However, in Model 3, for moderate cash flow sub-sample, results show weak
evidence of extreme earnings leading to incremental information content for moderate cash
flow from operations. The summed coefficients of the level and change of moderate cash flow
when earnings are extreme was positive and not significant based on cross-section regression,
and positive and significant at the 5 percent level based on the pooled-data regression.
The summed coefficients of the level and change of moderate cash flow when earnings are
extreme was positive and significant at the 1 percent level by using either the mean of the
yearly coefficients derived from yearly cross-section or the pooled-data regression when cash
flow (earnings) extremity is measured by cash flow (earnings) to price ratios (as reported in the
paper). These results reveal that measuring cash flow (earnings) extremity by cash flow
(earnings) to price ratios is superior to measuring cash flow (earnings) extremity by the
absolute value of changes in cash flow (earnings) scaled by beginning price for
detecting incremental information content for moderate cash when earnings are extreme.
18. This analysis is conducted for the mean of the summed yearly coefficients of the level and
change in earnings (cash flow) derived from annual regression. These results lead to the
same conclusion if the analysis is conducted for the summed coefficients of the level and
change in earnings (cash flow) in the case of the pooled regression.
19. Non-current accruals are less valued than both current accruals and cash flow from
operations. However, non-current accruals are considered informative component of
earnings (Rayburn, 1986; Bowen et al., 1987; Jennings, 1990; Pfeiffer et al., 1998; Pfeiffer and
Elgers, 1999).
20. Pfeiffer et al. (1998) reported a statistically significant differential valuation of cash flow and
current accruals when they employed a pooled cross-sectional and time-series data for
estimating surprises in earnings, working capital from operations, and cash flow from
operations.
References
Ali, A. (1994), “The incremental information content of earnings, working
capital from operations, and cash flows”, Journal of Accounting Research, Vol. 32 No. 1,
pp. 61-74.
Ali, A. and Pope, P.F. (1994), “The incremental information content of earnings, funds flow and
cash flow: the UK evidence”, working paper Version, The Management School, Lancaster
University, Lancaster, pp. 1-32.
Ali, A. and Pope, P.F. (1995), “The incremental information content of earnings, funds flow and cash
flow: the UK evidence”, Journal of Business Finance & Accounting, Vol. 22 No. 1, pp. 19-34.
Ali, A. and Zarowin, P. (1992a), “Permanent versus transitory components of annual earnings
and estimation error in earnings response coefficients”, Journal of Accounting and
Economics, Vol. 15 Nos 2/3, pp. 249-64.
Ali, A. and Zarowin, P. (1992b), “The role of earnings levels in annual earnings-returns studies”,
Journal of Accounting Research, Vol. 30 No. 2, pp. 286-96.
Arellano, M. (2003), “Unobserved heterogeneity”, Panel Data Econometrics: Advanced Texts in
Econometrics, Chapter 2, Oxford University Press, New York, NY.
Bahnson, P.R., Miller, P.B.W. and Budge, B.P. (1996), “Nonarticulation in cash flow statements
and implications for education, research and practice”, Accounting Horizons, Vol. 10 No. 4,
pp. 1-15.
Bernard, V.L. (1987), “Cross-sectional dependence and problems in inference in market-based
accounting research”, Journal of Accounting Research, Vol. 25 No. 1, pp. 1-48.
Bernard, V.L. and Stober, T.L. (1989), “The nature and amount of information in cash flows and
accruals”, The Accounting Review, Vol. 64 No. 4, pp. 624-52.
Biddle, G.C., Bowen, R.M. and Wallace, J.S. (1997), “Does EVA beat earnings? Evidence on
associations with stock returns and firm values”, Journal of Accounting and Economics,
Vol. 24 No. 3, pp. 301-36.
Biddle, G.C., Seow, G.S. and Siegel, A.F. (1994), “Relative versus incremental information
content”, Working Paper Version 1.1, University of Washington, Washington, DC,
September 1-28.
Biddle, G.C., Seow, G.S. and Siegel, A.F. (1995), “Relative versus incremental information
content”, Contemporary Accounting Research, Vol. 12 Nos 1/1, pp. 1-23.
Impact
of earnings
extremity
101
RAF
12,1
102
Board, J.L.G. and Day, J.F.S. (1989), “The information content of cash flow figures”, Accounting
& Business Research, Vol. 20 No. 77, pp. 3-11.
Board, J.L.G., Day, J.F.S. and Walker, M. (1989), The Information Content of Unexpected
Accounting Income, Funds Flow and Cash Flow, Institute of Chartered Accountants in
England and Wales, Edinburgh, pp. 1-16.
Bowen, R.M., Burgstahler, D. and Daley, L.A. (1987), “The incremental information content of
accrual versus cash flows”, The Accounting Review, Vol. 62 No. 4, pp. 723-47.
Brown, L.D., Hagerman, R.L., Griffin, P.A. and Zmijewski, M.E. (1987), “An evaluation of
alternative proxies for the market’s assessment of unexpected earnings”, Journal of
Accounting and Economics, Vol. 9 No. 2, pp. 159-93.
Brown, S.J. and Warner, J.B. (1980), “Measuring security price performance”, Journal of Financial
Economics, Vol. 8 No. 3, pp. 205-58.
Charitou, A., Clubb, C. and Andreou, A. (2001), “The effect of earnings performance, growth and
firm size on the usefulness of cash flows and earnings in explaining security returns:
empirical evidence for the UK”, Journal of Business Finance & Accounting, Vol. 28 Nos 5/6,
pp. 563-94.
Cheng, C.S.A. and Yang, S.S.M. (2003), “The incremental information content of earnings and
cash flows from operations affected by their extremity”, Journal of Business Finance
& Accounting, Vol. 30 Nos 1/2, pp. 73-116.
Cheng, C.S.A., Liu, C.-S. and Schaefer, T.F. (1996), “Earnings permanence and the incremental
information content of cash flows from operations”, Journal of Accounting Research,
Vol. 34 No. 1, pp. 173-81.
Cheng, C.S.A., Liu, C.-S. and Schaefer, T.F. (1997), “The value-relevance of SFAS No. 95 cash
flows from operations as assessed by security market effects”, Accounting Horizons,
Vol. 11 No. 3, pp. 1-15.
Christensen, P.O., Feltham, G.A. and Sabac, F. (2005), “A contracting perspective on earnings
quality”, Journal of Accounting and Economics, Vol. 39, pp. 265-94.
Clubb, C. (2003), “Discussion of the incremental information content of earnings and cash flows
from operations affected by their extremity”, Journal of Business Finance & Accounting,
Vol. 30 Nos 1/2, pp. 117-24.
Collins, D.W. and Kothari, S.P. (1989), “An analysis of intertemporal and cross-sectional
determinants of earnings response coefficients”, Journal of Accounting and Economics,
Vol. 11 Nos 2/3, pp. 143-81.
Das, S. and Lev, B. (1994), “Nonlinearity in the returns-earnings relation: tests of alternative
specifications and explanations”, Contemporary Accounting Research, Vol. 11 Nos 1/II,
pp. 353-79.
Donnelly, R. and Walker, M. (1995), “Share price anticipation of earnings and the effect of
earnings persistence and firm size”, Journal of Business Finance & Accounting, Vol. 22
No. 1, pp. 5-18.
Easton, P.D. and Harris, T.S. (1991), “Earnings as an explanatory variable for returns”, Journal of
Accounting Research, Vol. 29 No. 1, pp. 19-36.
EViews 5 User’s Guide (1994-2002), Quantitative Micro Software, LLC, Irvine, CA.
Fama, E.F. and Macbeth, J.D. (1973), “Risk return and equilibrium: empirical tests”, Journal of
Political Economy, Vol. 71, pp. 607-36.
Freeman, R.N. and Tse, S.Y. (1992), “A nonlinear model of security price responses to unexpected
earnings”, Journal of Accounting Research, Vol. 30 No. 2, pp. 185-209.
Green, J.P. (1999), “The impact of the quality of earnings on the valuation relevance of cash flow
disclosures”, British Accounting Review, Vol. 31 No. 4, pp. 387-413.
Greene, W.H. (2003), Econometrics Analysis, 5th ed., Pearson, Upper Saddle River, NJ.
Jennings, R. (1990), “A note on interpreting ‘incremental information content’”, The Accounting
Review, Vol. 65 No. 4, pp. 925-32.
Kennedy, P. (2003), A Guide to Econometrics, 5th ed., The MIT Press, Cambridge.
Kormendi, R. and Lipe, R. (1987), “Earnings innovations, earnings persistence, and stock
returns”, Journal of Business, Vol. 60 No. 3, pp. 323-45.
Kothari, S.P. (2001), “Capital markets research in accounting”, Journal of Accounting and
Economics, Vol. 31 Nos 1/3, pp. 105-231.
Leuz, C., Nanda, D. and Wysocki, P.D. (2003), “Earnings management and investor protection: an
international comparison”, Journal of Financial Economics, Vol. 69, pp. 505-27.
Maddala, G.S. (2001), Introduction to Econometrics, 3rd ed., Wiley, New York, NY.
Neill, J.D., Schaefer, T.F., Bahnson, P.R. and Bradbury, M.E. (1991), “The usefulness of
cash flow data: a review and synthesis”, Journal of Accounting Literature, Vol. 10,
pp. 117-50.
O’Hanlon, J., Poon, S. and Yaansah, R.A. (1992), “Market recognition of differences in earnings
persistence: UK evidence”, Journal of Business Finance & Accounting, Vol. 19 No. 4,
pp. 625-39.
Ou, J.A. and Penman, S.H. (1989), “Accounting measurement, price-earnings ratio, and
the information content of security prices”, Journal of Accounting Research, Vol. 27,
pp. 111-44.
Pfeiffer, J.R.J. and Elgers, P.T. (1999), “Controlling for lagged stock price responses in pricing
regressions: an application to the pricing of cash flows and accruals”, Journal of
Accounting Research, Vol. 37 No. 1, pp. 239-47.
Pfeiffer, J.R.J., Elgers, P.T., Lo, M.H. and Rees, L.L. (1998), “Additional evidence on the
incremental information content of cash flows and accruals: the impact of errors in
measuring market expectations”, The Accounting Review, Vol. 73 No. 3, pp. 373-85.
Rayburn, J. (1986), “The association of operating cash flow and accruals with security returns”,
Journal of Accounting Research, Vol. 24 No. 3, pp. 112-33.
Schaefer, T. and Kennelley, M. (1986), “Alternative cash flow measures and risk-adjusted
returns”, Journal of Accounting, Auditing & Finance, Vol. 1 No. 4, pp. 278-87.
Scott, W.R. (2003), Financial Accounting Theory, 3rd ed., Prentice-Hall, Toronto.
Wilson, G.P. (1986), “The relative information content of accruals and cash flows: combined
evidence at the earnings announcement and annual report release date”, Journal of
Accounting Research, Vol. 24 No. 3, pp. 165-200.
Wilson, G.P. (1987), “The incremental information content of the accrual and funds components
of earnings after controlling for earnings”, The Accounting Review, Vol. 62 No. 2,
pp. 293-322.
Wooldridge, J.M. (2002), Econometric Analysis of Cross Section and Panel Data, The MIT Press,
London, pp. 148-53.
About the authors
Wael Mostafa holds a PhD in Accounting and Finance from University of Durham, UK. He held
the position of Assistant Professor at College of Business Administration, University of
Bahrain. He has published several research papers in international referred journals such as
Impact
of earnings
extremity
103
RAF
12,1
104
International Research Journal of Finance and Economics and Middle Eastern Finance and
Economics. Wael Mostafa is the corresponding author and can be contacted at:
[email protected]
Professor Rob Dixon has been a member of staff in the Durham Business School at University
of Durham, UK, since 1992. He is the author of a number of books and articles in the broad field
of Management, and has published extensively in learned journals on the broad subject of
financial management and performance appraisal. He has published his research in Journal of
Accounting, Business and Management, Managerial Finance, Accounting Forum, Managerial
Auditing Journal, Accounting, Auditing and Accountability Journal. His current research includes
corporate finance, governance and control and related accounting issues.
To purchase reprints of this article please e-mail: [email protected]
Or visit our web site for further details: www.emeraldinsight.com/reprints