Description
Forecasting future period profitability is widely
identified as an aim of financial statement
analysis, and these forecasts are typically relied
upon for the estimation of firm value. To
facilitate this, the decomposition of earnings
into its components or drivers, is typically
advocated.
Accounting Research Journal
Earnings Decomposition and the Persistence of Earnings
Stephen Kean Peter Wells
Article information:
To cite this document:
Stephen Kean Peter Wells, (2007),"Earnings Decomposition and the Persistence of Earnings", Accounting Research
J ournal, Vol. 20 Iss 2 pp. 111 - 127
Permanent link to this document:
http://dx.doi.org/10.1108/10309610780000696
Downloaded on: 24 January 2016, At: 21:02 (PT)
References: this document contains references to 0 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 477 times since 2007*
Users who downloaded this article also downloaded:
David Heald, Ron Hodges, (2015),"Will “austerity” be a critical juncture in European public sector financial reporting?",
Accounting, Auditing & Accountability J ournal, Vol. 28 Iss 6 pp. 993-1015 http://dx.doi.org/10.1108/AAAJ -04-2014-1661
J oel D. Goldhar, David Lei, (1991),"The Shape of Twenty-first Century Global Manufacturing", J ournal of Business Strategy,
Vol. 12 Iss 2 pp. 37-41 http://dx.doi.org/10.1108/eb039401
Thomas Li-Ping Tang, J wa K. Kim, Theresa Li-Na Tang, (2002),"Endorsement of the money ethic, income, and life
satisfaction: A comparison of full-time employees, part-time employees, and non-employed university students", J ournal of
Managerial Psychology, Vol. 17 Iss 6 pp. 442-467 http://dx.doi.org/10.1108/02683940210439388
Access to this document was granted through an Emerald subscription provided by emerald-srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service
information about how to choose which publication to write for and submission guidelines are available for all. Please visit
www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio of
more than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of online
products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on Publication
Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.
*Related content and download information correct at time of download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
111
Earnings Decomposition and the
Persistence of Earnings
Stephen Kean and Peter Wells
School of Accounting
University of Technology
Abstract
Forecasting future period profitability is widely
identified as an aim of financial statement
analysis, and these forecasts are typically relied
upon for the estimation of firm value. To
facilitate this, the decomposition of earnings
into its components or drivers, is typically
advocated. This paper investigates the existence
of systematic differences in persistence across
the components of earnings. If components of
earnings experience differences in persistence,
this may provide insights into the determinants
of aggregate earnings level and persistence.
This paper provides evidence of differences in
persistence between components of earnings.
Differences are found between components
formed on the basis of: financial ratios;
operating and financing activities; and cash and
accruals. Furthermore, there is evidence that
earnings components improve the explanatory
power of models evaluating aggregate earnings
persistence, with this result being strongest for
firms with extreme income decreasing accruals.
Due to the pivotal role of earnings in firm
valuation, the results from this paper have direct
implications for valuation.
1. Introduction
The objective of this paper is to investigate
the existence of systematic differences in
persistence across the components of earnings,
and to determine whether these can be utilised
to improve the prediction of future period
performance. The bases for the decomposition
Acknowledgements: We appreciate the financial support of
the University of Technology, Sydney. Prior versions of this
paper have benefited from the comments of Alan Hodgson,
Anne Wyatt and participants at the 2006 AFAANZ
Conference and workshops at the University of Technology
Sydney.
Keywords: Financial reporting, earnings persistence.
of earnings evaluated include: component
financial ratios as suggested by financial
statement analysis texts such as Healy, Palepu
and Bernard (2004), Stickney, Brown and
Wahlen, (2005) and Penman (2007); operating
and financing activities (Fairfield, Sweeney and
Yohn, 1996); and cash and accrual components
(Sloan, 1996). Evidence is provided of
significant differences in persistence across
earnings components. There is also evidence of
the earnings components providing insights into
the persistence of earnings, with this result
being strongest for firms with extreme income
decreasing accruals.
When undertaking financial statement
analysis and business valuation, significant
efforts are expended on evaluating the historic
performance of the firm, and this is undertaken
with the aim of informing the forecasting of
future period performance. To facilitate this, the
decomposition of earnings into components is
commonly advocated. A primary motivation for
this paper is the determination whether earnings
decomposition aids the forecasting of future
period performance. Furthermore, if multiple
bases for decomposition are suggested, do
alternative decompositions better satisfy this
function? Earnings are the product of a range of
factors, including the application of accounting
practices (e.g., GAAP), the economic decisions
of the firm and its economic environment
(Francis, LaFond, Olsson & Schipper, 2004).
These factors have potentially distinct impacts
on the level and persistence of aggregate
earnings. Accordingly, analysis of the
persistence of alternative decompositions of
earnings, as suggested by Fairfield and Yohn
(2001), is undertaken.
Based on a sample of 2,845 ASX firm-year
observations over the period 1996-2004, evidence
is provided of significant differences in
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
112
persistence across earnings components,
including earnings decomposed based on:
financial ratios; operating and financing activities;
and cash and accrual components. Furthermore,
there is evidence of earnings components
improving the explanatory power of models
evaluating aggregate earnings persistence,
particularly for partitions of firms with extreme
income decreasing accruals. Such differences
support the decomposition of total earnings into
components for the purpose of persistence
analysis and financial statement analysis.
This paper contributes to the extant literature
as follows. First, earnings are the product of
mechanical accounting practices. Understanding
the operation of these practices, and in particular
how they affect the components of earnings,
provides insights into the persistence of
aggregate earnings. Second, earnings also reflect
management decision making. Understanding
how these manifest in earnings components
earnings also potentially provides insights into the
persistence of aggregate earnings.
The remainder of this paper is organised as
follows. Section 2 provides a review of the
existing literature and develops hypotheses.
Research design and sample selection are
addressed in sections 3 and 4. The results are
presented in section 5, and the conclusions and
suggestions for future research are discussed in
section 6.
2. Theory and Hypothesis
Development
The roles of earnings levels and persistence in
determining firm value are well recognised in
the literature. In analytical papers considering
the residual income model of firm valuation,
information on earnings level and persistence is
captured in either the determination of forecast
future period residual income, or as ‘other
information’ (Feltham & Ohlson, 1995; Ohlson,
1995). Further, there is empirical support for
earnings persistence influencing firm value with
Kormendi & Lipe (1987) and Easton &
Zmijewski (1989) finding more persistent
earnings eliciting more pronounced stock
returns at announcement date. The importance
of earnings persistence is identified in practice,
where it has a significant impact on the
determination of forecast future earnings and
appropriate terminal value assumptions (Nissim
& Penman, 2001). Reflecting this, considerable
attention has been devoted to evaluating
earnings persistence. For example, Dechow,
Hutton & Sloan (1999) considered a number
of model specifications to evaluate the extent
of earnings persistence, while Cheng
(2005) examined a range of ‘value-creating’
(economic) and ‘value-recording’ (accounting)
drivers of aggregate earnings persistence.
However, the persistence of the components
of earnings, and its implications for aggregate
earnings persistence has received relatively
limited attention. Forecasting future earnings
using current earnings involves making the
distinction between the permanent and
transitory nature of earnings. If certain
components of earnings are more permanent or
transitory than other components, then the
corresponding decomposition of earnings will
improve the forecasting of future earnings.
There is evidence of the decomposition of
earnings into profit margin and asset turnover
providing insights into future period earnings
(Fairfield and Yohn, 20001). Additionally, the
differing persistence of cash flows and accruals
has been considered (e.g., Sloan, 1996,
Fairfield, Whisenhant and Yohn, 2003),
although the primary concern here has been
evaluating the ‘accrual anomaly’ rather than the
use of components to explain future period
earnings. Accordingly, the issue addressed here
is whether there are systematic differences in
the persistence of earnings components, and
whether this can be exploited to improve the
prediction of future period earnings.
2.1 The decomposition of earnings into
financial ratio components
Financial statement analysis texts such as
Healy, Palepu and Bernard (2004) and Penman
(2007) typically advocate the decomposition of
earnings into its financial ratio components.
Financial ratios, the “building blocks of
forecasting,” relate financial statement
components to return on equity (Nissim &
Penman, 2001). This identifies the components
of earnings, or more specifically returns as
distinct, but complimentary, ‘drivers’ of future
earnings and generally takes the following
form:
1
1 For detailed decomposition of ROE into these
components refer to Penman (2007).
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
113
( ) [ ] ( )
t t t t t
SPREAD NFL ATO PM ROE * * + =
Where:
ROE
t
: Return on Equity for period t
PM
t
: Profit margin calculated as NOPAT
divided by sales for period t
NOPAT
t
: Net Operating Profit After Tax for
period t
ATO
t
: Asset turnover calculated as Sales
over NOA
ave
for period t
NOA
ave
: Average Net Operating Assets for
period i
NFL
i
: Leverage calculated as NFO
ave
divided by equity
ave
for period t
NFO
ave
: Average Net Financing Obligations
for period t
Equity
ave
: Average equity for period t
SPREAD
t
: Spread calculated as OROA –
NBC for period t
OROA
t
: Return on operating assets
calculated as NOPAT divided by
NOA
ave
for period t
NBC
t
: Net borrowing costs calculated as
NFE divided by NFO
ave
for period t
NFE
t
: Net Financing Expense, net of tax
shelter for period t
A feature of this approach is that it focuses
attention on the determinants of earnings and
captures information about both management
strategies and the economic environment within
which the firm operates.
Evidence on the persistence of the ratio
components of earnings is provided in Nissim
& Penman (2001), and it is inferred that forecast
return accuracy would be improved if earnings
were decomposed in this manner. However, the
portfolio approach adopted in evaluating
persistence limits the ability to demonstrate this
empirically. Some evidence of ratio
decomposition enhancing forecast return
accuracy is provided in Fairfield and Yohn
(2001). They find that decomposing change in
return on assets into change in asset turnover
and change in profit margin is useful in
forecasting next period change in return on
assets. However, there is no evidence of this
improving forecast accuracy on a levels basis.
This suggests further evaluation of the relation
between future period returns and lagged
financial ratio components of earnings.
2.2 The decomposition of earnings into
operating and financing components
A simpler approach to earnings decomposition
would be the classification of earnings by
nature. This approach is adopted in Fairfield
Sweeney and Yohn (1996), who decomposed
earnings into non-recurring, special, operating
and non-operating components. Supporting this
decomposition of earnings, they find
monotonically greater average explanatory
power (measured by adjusted r
2
) with greater
levels of disaggregation. Insights into the
determinants of this result are provided by
Hermann, Inoue & Thomas (2000) who find
significant differences in the persistence of such
components of earnings.
A similar decomposition of earnings is
suggested here, although a distinction between
operating and financing activities consistent
with Nissim & Penman (2001) is followed.
Modigliani & Miller (1958) suggest it is the
relatively uncertain returns from operating
activities of the firm that generate value, while
the financing activities have a zero net present
value. Due to the relatively fixed nature of
interest rates, Hermann, Inoue & Thomas
(2000) suggest that financing income is more
persistent than operating income. Similarly, by
their non-recurring nature, ‘other’ accounting
items
2
will be less persistent (more transitory)
than operating income.
2.3 The decomposition of earnings into cash
and accrual components
Finally, many studies have decomposed
earnings into cash flows and accruals, and
considered the information conveyed by the
separate components. For example, Dechow
(1994) examined the extent to which cash flows
and accruals capture firm performance, finding
that accruals enhanced income numbers as
measures of firm performance. This result was
more pronounced for firms with more volatile
cash flows. Similarly, Sloan (1996) provides
evidence of cash flow components of earnings
being more persistent than accruals, and
contrary to expectation this appears not to be
fully reflected in share prices. Subsequent
studies have further decomposed cash flow and
2 For example, abnormal-items, extraordinary-items, gain-
on-sale of non-current assets, and gains on foreign
currency transactions.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
114
accrual components of earnings (e.g., Xie,
2001; Dechow & Ge, 2005; Richardson, Sloan,
Soliman & Tuna, 2005, Zhang, 2007).
However, these consider the relevance of
earnings components relative to share prices,
rather than evaluating earnings persistence and
whether it benefits forecasting. Sloan (1996)
suggests some accruals are inherently more
transitory than other accruals, and cash flows
are less vulnerable to distortion than accruals.
Fairfield et al (2003) suggest the lower
persistence of accruals reflects accounting
conservatism and the diminishing marginal
return on new project investments. Richardson,
Sloan, Soliman & Tuna (2005) highlight the
subjectivity in the estimation of accruals and a
greater vulnerability to intentional and
unintentional accounting distortions.
Importantly, the tendency for accruals to reverse
is identified, and this is reflected in models of
the persistence of accruals. However,
unresolved is whether separation of earnings
into cash flow and accrual components will
better explain future period earnings.
3. Research Design
3.1 Establishing a benchmark
Central to the evaluation of whether the
decomposition of earnings improves the
evaluation of earnings persistence is the
determination of an appropriate benchmark. To
avoid scale problems earnings are scaled by
equity, and an autoregressive model of earnings
scaled by book value of equity (i.e., ROE) is
utilised. In the literature a number of variations
of book value of equity are used (i.e., opening,
average and closing), and consideration is first
given to the determination of the variation to be
used in this study.
3
Accordingly the following
model is estimated:
(1)
where:
ROE
it
:
it
it
BVEquity
Earnings
3 Research examining earnings persistence often utilises
lagged Equity
i,t-1
(e.g. Ohlson 1995, Fairfield et al 1996,
Dechow et al 1999 Herrmann et al 2000, Penman &
Zhang 2002, Cheng 2005), possibly to avoid the
endogenous effect of Earnings
i,t
upon Equity
i,t
. However,
Sloan (1996), Fairfield & Yohn, (2001) and Nissim &
Penman (2001) utilise average equity to minimise the
effect of substantial changes in equity through the period.
Earnings
it
: Net Profit After Tax Before
Outside Equity Interests for firm i
in year t
BVEquity
it
: Book Value of equity for firm i in
year t, measured as either
i. Opening value
ii. Average Value
iii. Closing Value
The intercept, ?
0
, represents the average level
of ROE independent of lagged ROE, while the
slope coefficient, ?
1
, represents the average
persistence (or reversion toward zero) of ROE.
4
3.2 Persistence of components of earnings
An initial stage in the evaluation of differences
in the persistence of components of earnings is
the estimation of an autoregressive model for
each of the components. This takes the same
general form as above, and is as follows:
(2)
where Component is represented by the
following variables
Financial Ratio Components
PM
it
:
it
it
Sales
NOPAT
ATO
it
:
1 ? it
it
NOA
Sales
NFL
it
:
1
1
?
?
it
it
Equity
NFO
SPREAD
it
:
( )
it it it
NBC ATO PM ? *
Operating, Financing and Other Components
OROE
it
: Earnings from Operating activities
scaled by lagged book value of
equity, for firm i in year t
FROE
it
: Earnings from Financing activities
scaled by lagged book value of
equity, for firm i in year t
OTHER
it
: Other income scaled by lagged
book value of equity, for firm i in
year t
Cash and Accruals Components
CROE
it
: Cash Flow from Operations scaled
by lagged book value of equity, for
firm i in year t
4 Neither equation (1) nor later equations contain RHS
variables other than earnings and its components. This
reflects the concern in this type of study with variation in
persistence of earnings components, rather than variation
in persistence across firms.
ROE
it
= ?
0
+ ?
1
ROE
it-1
+?
it
Component
it
= ?
0
+ ?
1
Component
it-1
+ ?
it
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
115
AROE
it
: Total Accruals scaled by lagged
book value of equity, for firm i in
year t
Differences in the persistence of the
components of earnings will manifest in
differences in model explanatory power, and co-
efficients on the component lagged values.
3.3 Components of earnings and the
persistence of aggregate earnings
To evaluate whether the decomposition of
earnings provides additional information
relevant to the evaluation of earnings
persistence, the following model is estimated
for the different earnings decompositions (i.e.,
financial ratios; operating, financing and other;
and cash and accruals):
(3)
Whether differences in the persistence of
components of earnings can be exploited to
enhance the estimation of future period earnings
will be identified by improvements in model
explanatory power.
4. Data Collection and Sample
Description
4.1 Data collection
Financial data for companies on the ASX from
1996-2004 was collected from the Aspect-
Huntley database. For inclusion, firm-year
observations are required to have all available
data for two consecutive years of persistence
calculation. Observations with earnings-based
ratios beyond +/-1.00 and balance-sheet-based
ratios above 5.00 were excluded to control for
the effect of outliers. Nissim & Penman (2001)
document relatively low persistence for extreme
deciles of ratios, and thus the exclusion of
extreme-ratio observations will bias the
persistence coefficients upwards. Due to
complications in applying the Nissim &
Penman (2001) methodology of removing the
effect of leverage from operating performance
for financial services firms, firms in the
financial services sector are excluded from the
sample.
The sample selection processes, and the
impact of deletions are reported in Table 1. This
provided a final sample of 2,854 firm-year
observations. This highlights the survivorship
bias that is introduced by the requirement for
three years of financial data, and this is
doubtless exacerbated by the exclusion of
observations with extreme ratios. The number
of observations by industry and year was
considered, and no year appears to dominate the
sample, although consistent with Australian
market characteristics the Materials sector does
comprise 23% of the sample.
4.2 Sample description
Descriptive statistics on the variables are
provided in Table 2. It is notable that 22% of
the sample has negative total earnings (ROE).
While this is more than twice the proportion of
losses in Dechow (1994), which was based
on a sample of U.S. firms over the period
1960–1989, it is less than Watson and Wells
(2005). This is likely a consequence of the
sample selection bias discussed above. Poor
performance is not limited to earnings (ROE),
17% of the sample has negative operating
earnings (OROE and PM), and 15% have
negative cash flow from operations (CROE).
The sample mean for ‘Other’ items was also
Table 1
Sample Selection
Firms Firm-years
All firms included in Aspect database, 1996-2004 2,097 14,560
Less: Financial Services firms 420 2,826
Less: Missing data and IPOs 76 529
Less: Foreign currency reporting firms 289 3,214
Less: Observations with at least one extreme ratio* 628 5,146
Final sample size 1996–2004 684 2,845
* Excluded firms: Sales ? 0, Equity ? 0, Operating Assets ? 0, or Net Operating Assets ? 0, -1>Earnings based ratio’s>1,
0?Balance Sheet based ratio>5
ROE
it
= ?
0
+ ??
j
Component
it-1
+ ?
it
k
j=1
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
116
Table 2
Descriptive Statistics
Variable Mean Median Std. Dev. Minimum Maximum Negative
ROE 0.076 0.090 0.180 -0.888 0.936 22%
RNOA 0.08 0.08 0.16 -0.91 1.00 16%
RLOS -0.01 0.00 0.12 -0.84 0.84 44%
PM 0.044 0.050 0.163 -0.983 0.978 17%
ATO 2.036 1.543 1.698 0.010 9.989 0%
NFL 0.348 0.294 0.634 -0.937 7.076 28%
SPREAD 0.031 0.031 0.269 -0.998 0.995 39%
OROE 0.109 0.115 0.168 -0.888 0.975 17%
FROE -0.020 -0.017 0.042 -0.362 0.827 75%
OTHER -0.013 0.000 0.086 -0.892 0.952 41%
CROE 0.169 0.160 0.213 -0.982 1.000 15%
AROE -0.094 -0.084 0.192 -0.984 0.976 76%
Where:
ROE Return-on-Start-of-Period-Equity Spread OROA-NBC
RNOA Return-on-Net-Operating-Assets OROE Operating-Return-on-Equity
RLOS Return-on-Levered-Operating-Spread FROE Financing-Return-on-Equity
PM Profit-Margin OTHER Other-Return-on-Equity
ATO Asset-Turnover CROE Cash-Return-on-Equity
NFL Leverage AROE Accrual-Return-on-Equity
negative, with 41% of the sample observations
being negative which is consistent with the
increasing incidence of firms reporting negative
unusual items. This has previously been
documented in Elliott & Hanna (1996), and is
possibly motivated by the smoothing of ‘usual’
earnings (Francis et al, 2004). This volatility of
earnings, which may be increasing, is likely to
be problematic when evaluating earnings
persistence.
Spearman and Pearson correlations were
calculated (untabulated) and most of the
variables are significantly correlated using both
parametric and non-parametric tests. This broad
correlation amongst the variables suggests a
substantial multicollinearity problem, and is
common in earnings component analysis
(e.g. Nissim & Penman, 2001).
5. Results
5.1 Establishing a benchmark
Central to the evaluation of how earnings
decomposition provides insights into earnings
persistence is the determination of an
appropriate benchmark. Three alternative
specifications of ROE are considered and the
results (untabulated) are not materially different.
As ROE determined by dividing earnings by
beginning of period equity has the highest
t-statistic on ?
1,
and adjusted R
2
(?
1
=0.486,
t-stat=29.077 and adjusted R
2
=0.229) and lacks
an endogeneity issue arising from the inclusion
of current period earning in the measure of
equity, this paper uses beginning of period
equity as the denominator for all scaled
variables.
However, it is notable that the co-efficient on
lagged ROE (?
1
) is materially lower than that
reported by Sloan (1996), only 0.486 compared
to 0.841.
5
Furthermore, the model has relatively
low explanatory power. This suggests much
lower earnings persistence in the present
sample. To gain insights into the magnitude of
the problem, firms were partitioned into
quintiles based on ROE in period t-1, and this
was compared with their quintile ranking in the
subsequent period. The results are presented in
Table 3, and shows considerable volatility in
quintile membership, with as few as 35.5% of
firms maintaining their quintile ranking across
5 Sloan (1996) utilised operating income scaled by total
assets, not ROE.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
117
Table 3
ROE Persistence — Changes in Quintile Ranking from Period t-1 to t
Evaluation of the stability of ROE between periods by considering differences in firm quintile rankings
across periods
Profit 5 8.8% 4.4% 8.3% 22.8% 55.7%
4 6.9% 7.6% 23.9% 41.3% 20.4%
3 10.4% 26.4% 35.5% 18.5% 9.3%
ROE
i,t
2 24.1% 36.5% 21.6% 10.9% 6.9%
Loss 1 49.8% 25.1% 10.7% 6.5% 7.7%
100% 100% 100% 100% 100%
Quintile 1 2 3 4 5
Loss ROE
i,t-1
Profit
Where:
ROEit :
1 ? it
it
BVEquity
Earnings
Earnings
it
: Net Profit After Tax Before Outside Equity Interests for firm I in year t
Table 4
ROE Persistence within Accrual Quintiles
Evaluation of the persistence of ROE for both the full sample of firms (Panel A) and for firms partitioned
on the basis of accruals relative to cash flow (Panel B)
Panel A: Persistence for Full Sample
(n=2,845) Coeff. t-stat Adjusted R
2
Intercept 0.032 10.114
***
0.229
ROE
i,t-1
0.486
29.077
***
Panel B: Persistence Within Accrual Quintiles
Portfolios constructed on the basis of firms ranked by lagged accruals scaled by lagged absolute cash flow
from operations
Coeff. t-stat Adjusted R
2
Quintile 1 Intercept
ROE
i,t-1
-0.020
0.115
-1.936
2.051
**
0.006
Quintile 2 Intercept
ROE
i,t-1
0.019
0.698
2.366
**
10.283
***
0.156
Quintile 3 Intercept
ROE
i,t-1
0.035
0.640
4.720
***
15.647
***
0.300
Quintile 4 Intercept
ROE
i,t-1
0.019
0.641
2.305
**
18.126
***
0.366
Quintile 5 Intercept
ROE
i,t-1
0.006
0.439
0.666
10.250
***
0.155
(1)
All variables as previously defined
*** Two tailed t-test significant at 1% level
** Two tailed t-test significant at 5% level
* Two tailed t-test significant at 10% level
ROE
it
= ?
0
+ ?
1
ROE
it-1
+ ?
it
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
118
the two periods. For example, 32.2% of lagged
quintile 3 firms increased their quintile ranking,
and 8.3% of firms are ranked in quintile 5 in the
subsequent period. Similarly 32.3% of firms
had a lower quintile ranking, and 10.7% of
firms are ranked in quintile 1 in the subsequent
period.
Prior research has identified earnings
persistence as being sensitive to growth (e.g.,
Fairfield, Whisenant and Yohn, 2003), and
large values of negative accruals have been
attributed to unusual items or write-downs
(Elliot and Hanna, 1996). Extreme negative (or
positive) accruals are likely to be transitory.
Accordingly, model 1 is re-estimated with firms
partitioned into quintiles based on lagged
accruals relative to lagged absolute cash flows.
The results are presented in Table 4 and this
identifies considerable variation across the
partitions. For firms experiencing extreme
income decreasing accruals (i.e., quintile 1)
ROE has relatively low persistence (?
1
=0.115, t-
stat=2.051, adjusted r
2
=0.006). While for firms
with income increasing accruals (i.e., quintile
5), there is more persistence in ROE (?
1
=0.439,
t-stat=10.250, adjusted r
2
=0.155). However, the
persistence is much higher for the middle
quintiles, and in particular quintile 4 (?
1
=0.641,
t-stat=18.126, adjusted r
2
=0.366).
6
Evidence in
Watson and Wells (2005) also suggests that
there is a much greater association between
earnings and stock price for these firms.
Accordingly, for firms where reported earnings
better captures underlying firm performance,
earnings persistence is higher. This confirms the
undertaking of subsequent tests on firms
partitioned in this manner.
5.2 Persistence of components of earnings
Attention is directed initially to the evaluation
of the persistence of components of earnings,
and the results of estimating equation 2 are
presented in Table 5. Panel A reports the results
for financial ratio components of earnings. For
the full sample, the co-efficient on profit margin
(PM) is 0.578, while the adjusted r
2
is 0.324.
Notably, there is considerable variation across
firms partitioned on the basis of lagged accruals.
There is much greater persistence in quintiles 3
and 4, measured by the co-efficient on profit
6 For quintile 3 the results of ?
1
=0.640, t-stat=15.647 and
adjusted r
2
=0.300 are similar.
margin (0.693 and 0.676 respectively), and
adjusted r
2
(0.461 and 0.491 respectively) than
in the remaining quintiles (e.g., for quintile 1,
?
1
=0.412, adjusted r
2
=0.166 and quintile 5,
?
1
=0.547, adjusted r
2
=0.218).
This contrasts with the results for asset
turnover (ATO) and leverage (NFL) which
suggest relatively more persistence across all
firms. This is captured in the regressions based
on the full sample of firms, with the co-efficient
on asset turnover (ATO) being 0.797 and the
adjusted r
2
is 0.678, while the co-efficient on
leverage (NFL) is 0.663 and the adjusted r
2
is
0.522. Finally, there is little persistence in the
margin between operating returns and
borrowing costs (SPREAD), with an adjusted r
2
for the full sample of only 0.059, and a
maximum adjusted r
2
of 0.098 for quintile 4.
In combination these suggest considerable
persistence in asset turnover and leverage across
all firms, while profit margin is most persistent
across quintile 3 and 4 firms. In contrast, for the
margin between operating returens and
borrowing costs there is little persistence.
Panel B reports the results for the operating,
financing and other components of earnings.
For the full sample the co-efficient on operating
returns (OROE) is 0.530 and the adjusted r
2
is
0.288. However, there is considerable variation
across the quintiles. For quintiles 2 to 4 there is
greater persistence in operating returns
(OROE), measured by the co-efficient ?
1
(0.733, 0.538 and 0.603 respectively) and
adjusted r
2
(0.25, 0.315 and 0.408 respectively).
However, there is considerable variation across
quintiles 1 and 5 (for example, in quintile 1
?
1
=0.342 and adjusted r
2
=0.081, and in quintile
5 ?
1
=0.465 and adjusted r
2
=0.226). In contrast
there is less variation in the persistence of
financing returns (FROE) across the quintiles
(?
1
=0.326 to 0.623 and adjusted r
2
=0.157 to
0.379), and little persistence in OTHER with a
maximum adjusted r
2
of 0.020 recorded.
Finally the persistence of cash and accrual
components of earnings is presented in Panel C.
There is little variation in the coefficients on
cash returns (CROE) across quintiles 2 to 4
(?
1
=0.514 to 0.532), and the models exhibit
similar explanatory power (adjusted r
2
=0.249 to
0.278). In contrast, for quintile 1 and 5 firms,
the persistence of cash returns (CROE) is much
lower (?
1
=0.322 and 0.368 respectively), and
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
119
the model explanatory power is much reduced
(adjusted r
2
=0.063 and 0.089 respectively).
With respect to accruals, these are subject to
much lower persistence with a maximum
adjusted r
2
of only 0.082. However, consistent
with the results above, there are not major
differences across quintiles 2 to 4 (?
1
=0.290 to
0.368), which contrasts with the results for
quintile 1 and 5 firms (?
1
=0.100 and 0.041
respectively)
In summary, across all the earnings
decompositions there is considerable variation in
the persistence of the components of earnings,
measured by either the co-efficient on the lagged
component or the explanatory power of the
models. Furthermore, these differences are more
pronounced with firms partitioned on the basis of
accruals relative to cash flows. This raises the
question of whether these differences can be
relied upon to better explain earnings persistence.
Table 5
Persistence of Components of Earnings Within Quintiles of
Firms Partitioned on the Basis of Accruals
Evaluation of the persistence of components of earnings for the full sample of firms and for firms
partitioned on the basis of lagged accruals relative to cash flow. This is undertaken for components
determined on the basis of: financial ratios (Panel A); operating, financing and other activities (Panel B);
and cash and accrual components (Panel C).
Panel A: Persistence of Financial Ratio Components
Coeff. t-stat Adjusted R
2
Full Sample Intercept
PM
i,t-1
0.017
0.578
6.673
***
36.915
***
0.324
Quintile 1 Intercept
PM
i,t-1
-0.008
0.412
-1.067
***
10.664
0.166
Quintile 2 Intercept
PM
i,t-1
0.019
0.625
3.713
***
15.897
***
0.307
Quintile 3 Intercept
PM
i,t-1
0.022
0.693
4.883
***
22.053
***
0.461
Quintile 4 Intercept
PM
i,t-1
0.018
0.676
3.741
***
23.452
***
0.491
Quintile 5 Intercept
PM
i,t-1
0.003
0.547
0.466
12.622
***
0.218
Full Sample Intercept
ATO
i,t-1
0.349
0.797
12.913
***
77.421
***
0.678
Quintile 1 Intercept
ATO
i,t-1
0.500
0.740
7.603
***
24.192
***
0.507
Quintile 2 Intercept
ATO
i,t-1
0.086
1.000
1.780
*
47.147
***
0.796
Quintile 3 Intercept
ATO
i,t-1
0.207
0.893
3.888
***
44.510
***
0.777
Quintile 4 Intercept
ATO
i,t-1
0.366
0.779
5.776
***
35.596
***
0.690
Quintile 5 Intercept
ATO
i,t-1
0.406
0.692
6.264
***
32.373
***
0.648
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
120
Table 5 (cont.)
Persistence of Components of Earnings Within Quintiles of
Firms Partitioned on the Basis of Accruals
Panel A: Persistence of Financial Ratio Components of Earnings
Coeff. t-stat Adjusted R
2
Full Sample Intercept
NFL
i,t-1
0.126
0.663
14.449
***
55.725
***
0.522
Quintile 1 Intercept
NFL
i,t-1
0.167
0.716
6.738
***
22.804
***
0.542
Quintile 2 Intercept
NFL
i,t-1
0.077
0.762
4.821
***
33.868
***
0.506
Quintile 3 Intercept
NFL
i,t-1
0.112
0.568
6.340
***
21.380
***
0.445
Quintile 4 Intercept
NFL
i,t-1
0.087
0.723
5.259
***
24.158
***
0.506
Quintile 5 Intercept
NFL
i,t-1
0.163
0.599
7.800
***
25.920
***
0.477
Full Sample Intercept
SPREAD
i,t-1
0.021
0.246
4.337
***
13.358
***
0.059
Quintile 1 Intercept
SPREAD
i,t-1
-0.047
0.080
-3.657
***
2.015
**
0.005
Quintile 2 Intercept
SPREAD
i,t-1
0.018
0.229
1.942
*
4.811
***
0.038
Quintile 3 Intercept
SPREAD
i,t-1
0.038
0.357
3.738
***
7.738
***
0.094
Quintile 4 Intercept
SPREAD
i,t-1
0.038
0.332
3.336
***
7.904
***
0.098
Quintile 5 Intercept
SPREAD
i,t-1
0.024
0.163
1.980
*
3.719
***
0.022
Panel B: Persistence of Operating, Financing and Other Components of Earnings
Coeff. t-stat Adjusted R
2
Full Sample Intercept
OROE
i,t-1
0.046
0.530
14.667
***
33.904
***
0.288
Quintile 1 Intercept
OROE
i,t-1
0.029
0.342
3.905
***
7.137
***
0.081
Quintile 2 Intercept
OROE
i,t-1
0.025
0.733
3.163
***
13.794
***
0.250
Quintile 3 Intercept
OROE
i,t-1
0.064
0.538
9.230
***
16.193
***
0.315
Quintile 4 Intercept
OROE
i,t-1
0.043
0.603
5.909
***
19.811
***
0.408
Quintile 5 Intercept
OROE
i,t-1
0.034
0.465
4.187
***
12.905
***
0.226
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
121
Table 5 (cont.)
Persistence of Components of Earnings Within Quintiles of
Firms Partitioned on the Basis of Accruals
Panel B: Persistence of Operating, Financing and Other Components of Earnings
Coeff. t-stat Adjusted R
2
Full Sample Intercept
FROE
i,t-1
-0.012
0.435
-16.321
***
27.793
***
0.213
Quintile 1 Intercept
FROE
i,t-1
-0.017
0.374
-10.626
***
11.793
***
0.196
Quintile 2 Intercept
FROE
i,t-1
-0.017
0.623
-3.055
**
11.866
***
0.198
Quintile 3 Intercept
FROE
i,t-1
-0.014
0.326
-9.588
***
11.527
***
0.188
Quintile 4 Intercept
FROE
i,t-1
-0.010
0.355
-7.661
***
10.330
***
0.157
Quintile 5 Intercept
FROE
i,t-1
-0.009
0.556
-6.287
***
18.626
***
0.379
Full Sample Intercept
Other
i,t-1
-0.012
0.117
-7.800
***
6.515
***
0.014
Quintile 1 Intercept
Other
i,t-1
-0.023
0.090
-4.016
***
2.421
**
0.008
Quintile 2 Intercept
Other
i,t-1
-0.013
-0.130
-4.464
***
-1.864
*
0.004
Quintile 3 Intercept
Other
i,t-1
-0.005
0.032
-2.227
**
0.555
0.000
Quintile 4 Intercept
Other
i,t-1
-0.009
0.179
-2.994
***
3.509
***
0.020
Quintile 5 Intercept
Other
i,t-1
-0.016
0.143
-4.519
***
3.431
***
0.019
Panel C: Persistence of Cash Flow and Accruals Components of Earnings
Coeff. t-stat Adjusted R
2
Full Sample Intercept
CROE
i,t-1
0.099
0.447
21.833
***
26.800
***
0.201
Quintile 1 Intercept
CROE
i,t-1
0.079
0.322
6.522
***
6.280
***
0.063
Quintile 2 Intercept
CROE
i,t-1
0.075
0.531
6.221
***
14.574
***
0.271
Quintile 3 Intercept
CROE
i,t-1
0.098
0.514
9.141
***
14.838
***
0.278
Quintile 4 Intercept
CROE
i,t-1
0.086
0.532
8.307
***
13.748
***
0.249
Quintile 5 Intercept
CROE
i,t-1
0.114
0.368
12.973
***
7.526
***
0.089
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
122
Table 5 (cont.)
Persistence of Components of Earnings Within Quintiles of
Firms Partitioned on the Basis of Accruals
Panel C: Persistence of Cash Flow and Accruals Components of Earnings
Coeff. t-stat Adjusted R
2
Full Sample Intercept
AROE
i,t-1
-0.082
0.238
-21.662
***
13.287
***
0.058
Quintile 1 Intercept
AROE
i,t-1
-0.133
0.100
-7.712
***
1.834
*
0.004
Quintile 2 Intercept
AROE
i,t-1
-0.074
0.368
-6.364
***
7.208
***
0.082
Quintile 3 Intercept
AROE
i,t-1
-0.071
0.290
-6.523
***
3.650
***
0.021
Quintile 4 Intercept
AROE
i,t-1
-0.057
0.304
-6.504
***
1.924
*
0.005
Quintile 5 Intercept
AROE
i,t-1
-0.046
0.041
-3.725
***
0.730
0.000
(2)
All variables as previously defined
*** Two tailed t-test significant at 1% level
** Two tailed t-test significant at 5% level
* Two tailed t-test significant at 10% level
5.3 Components of earnings and the
persistence of aggregate earnings
Evidence of whether the decomposition of
earnings into components better explains the
persistence of aggregate earnings is presented in
Table 6, and attention is directed in the first
instance towards the decomposition of earnings
into financial ratio components. When lagged
financial ratio components (PM, ATO, FLEV
and SPREAD) are regressed on ROE the
adjusted r
2
declines to 0.157. However, an issue
with this analysis is that it assumes an additive
relation between the components of earnings,
which is not the case for these financial ratios,
and equation 3 would be misspecified.
Addressing this problems, the financial ratio
components are partially aggregated to provide
a linear model (i.e., RNOA = PM *ATO and
RLOS = FLEV*SPREAD) and regressed on
ROE. The adjusted r
2
for this model increases to
0.244, which confirms the original model as
being poorly specified. Importantly, the co-
efficients on the return on net operating assets
(RNOA) and return attributable to leverage
(RLOS) are significantly different and the
increase in adjusted r
2
is significant (Vuong test
z stat = 20.430). Accordingly, there is evidence
of financial ratio components of earnings
exhibiting differing persistence, and evidence of
this being able to be exploited to provide
insights into the persistence of earnings.
Attention now shifts to the decomposition
of earnings operating, financing and other
components. The co-efficients on operating
and financing components of earnings, OROE
t-1
(?
1
= 0.565, t-stat = 30.408) and FROE
t-1
(?
2
= 0.602, t-stat = 8.635) are not significantly
different from each other (F-stat = 0.317).
However, they are significantly different from
the co-efficient on OTHER
t-1
(?
4
= 0.224,
t-stat = 6.782, F-stat = 84.944). Hence, there is
limited of these components of earnings having
differing persistence. Exploiting this limited
difference in persistence, the overall explanatory
power of the model increased, with the adjusted
R
2
increasing from 0.229 to 0.251 (i.e., an
increase of almost 10%), which is statistically
significant (Vuong test z stat = 20.273).
Consequently, there is again limited evidence of
these components of earnings exhibiting
difference persistence, and this being able to be
exploited to better explain earnings persistence.
Component
it
= ?
0
+ ?
1
Component
it-1
+ ?
it
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
123
Table 6
Earnings Components and the Persistence of Earnings
Evaluation of whether the decomposition of earnings into components increases the explanatory power of
models of earnings persistence.
ROE (Benchmark) Financial Ratios A Financial Ratios B
Operating, Financing
and Other
Cash Flow and
Accruals
Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat
Intercept 0.032 10.114
***
0.018 5.450
***
0.023 4.297
***
0.022 6.126
***
0.014 3.747
***
ROE
i,t-1
0.486 29.077
***
RNOA
i,t-1
0.578 30.245
***
PM
i,t-1
0.254 10.868
***
ATO
i,t-1
0.012 6.350***
RLOS
i,t-1
0.441 13.661
***
NFL
i,t-1
0.015 3.220
***
SPREAD
i,t-1
0.179 10.001
***
OROE
i,t-1
0.565 30.408
***
FROE
i,t-1
0.602 8.635
***
Other
i,t-1
0.224 6.782
***
CROE
i,t-1
0.545 30.954
***
AROE
i,t-1
0.396 20.819
***
N 2845 2845 2845 2845 2845
Adjusted R
2
0.229 0.244 0.157 0.251 0.252
F-stat 845.444
***
459.581
***
118.554
***
318.479
***
480.391
***
(3)
All variables as previously defined
*** Two tailed t-test significant at 1% level
** Two tailed t-test significant at 5% level
* Two tailed t-test significant at 10% level
Finally attention is directed to the
decomposition of earnings into cash and accrual
components. The co-efficients on the cash
component of ROE (?
1
= 0.545, t-stat = 30.954)
and the accrual component of ROE
(?
2
= 0.396,
t-stat = 20.819) are significantly different
(F-stat = 89.129). Hence, there is evidence of
these components of earnings having differing
persistence. Further, the overall explanatory
power of the model increased because of this
decomposition of earnings, with the adjusted R
2
increasing from 0.229 to 0.252 (i.e., an increase
of 10%) which is statistically significant
(Vuong test z stat = 20.377).
Accordingly, there is evidence of these
components of earnings exhibiting differences
in persistence, and this being able to be
exploited to better explain earnings persistence.
5.4 Components of earnings and the
persistence of aggregate earnings by quintiles
Whether earnings decomposition provides
insights into earnings persistence may be
influenced by observations where there is little
persistence in earnings, or where the stability of
earnings masks the benefits of decomposition.
Accordingly, tests of whether the decomposition
of earnings into components better explains the
persistence of aggregate earnings is undertaken
separately for quintiles based on lagged accruals
relative to lagged absolute cash flows.
The results for earnings decomposed into
financial ratio components are presented in
Table 7, Panel A.
7
There is considerable
variation in explanatory power of the model
7 Results for regression of lagged PM, ATO, FLEV and
SPREAD on ROE are not reported due to the previously
identified problem with model specification.
ROE
it
= ?
0
+ ? ?
j
Component
it-1
+ ?
it
k
j=1
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
124
across the quintiles, with the adjusted r
2
ranging
from 0.020 to 0.371. Comparison of the results
with the benchmarks in Table 4 suggests that
only for quintile 1 firms are there benefits from
decomposition (i.e., adjusted r
2
increasing from
.006 to 0.020), and this partition of firms is
likely driving the result in Table 6.
The results for earnings decomposed into
operating, financing and other components are
presented in Panel B. Consistent with prior
results this shows there is considerable variation
in explanatory of the model across the quintiles,
with the adjusted r
2
ranging from 0.024 to
0.368. Comparison of the results with the
benchmarks in Table 4 again suggests that there
are few benefits to this decomposition of
earnings for quintile 3, 4 and 5 firms, with only
minimal increases in adjusted r
2
(0.300 v 0.302,
0.366 v 0.368 and 0.155 v 0.163). The benefits
of decomposition appear largest for quintile 1
and 2 firms that are reporting relatively large
negative accruals where the adjusted r
2
increases (0.006 v 0.024 and 0.156 v 0.231),
although the models continue to suffer from
relative low explanatory power.
The results are generally weaker for earnings
decomposed into cash flows and accruals
components, with the results presented in Panel
C. Again, there is considerable variation in
explanatory power across the quintiles with the
adjusted r
2
ranging from 0.018 to 0.370.
However, comparison of these results with the
benchmarks in Table 4 shows that only for
quintile 1 firms are there limited benefits from
decomposition (i.e., adjusted r
2
increasing from
.006 to 0.018).
In summary, from Table 7 it is apparent that
where there is less persistence in ROE,
particularly quintile 1 firms there is some
support of the decomposition of earnings
providing insights into earnings persistence and
this is driving the result in Table 6. However,
for remaining firms the decomposition of
earnings into components provides only limited
insights in earnings persistence that are not
provided by earnings generally.
Table 7
Components of Earnings and the Persistence of Earnings Within
Accrual Quintiles
Evaluation of the decomposition of earnings provides insights into aggregate earnings persistence for the
full sample of firms and for firms partitioned on the basis of lagged accruals relative to cash flow. This is
undertaken for earnings decomposed on the basis of: financial ratios (Panel A); operating and financing and
other activities (Panel B); and cash and accrual components (Panel C).
Panel A: Financial Ratio Components
Coeff. t-stat Adjusted R
2
Full Sample Intercept
RNOA
i,t-1
RLOS
i,t-1
0.018
0.578
0.441
5.450
***
30.245
***
13.661
***
0.244
Quintile 1 Intercept
RNOA
i,t-1
RLOS
i,t-1
-0.026
0.201
0.011
-2.421
**
3.238
***
0.165
0.020
Quintile 2 Intercept
RNOA
i,t-1
RLOS
i,t-1
0.012
0.770
0.533
1.527
***
10.727
***
6.052
***
0.167
Quintile 3 Intercept
RNOA
i,t-1
RLOS
i,t-1
0.034
0.652
0.564
4.665
***
15.757
***
9.462
***
0.303
Quintile 4 Intercept
RNOA
i,t-1
RLOS
i,t-1
0.019
0.649
0.529
2.330
**
18.348
***
9.063
***
0.371
Quintile 5 Intercept
RNOA
i,t-1
RLOS
i,t-1
0.003
0.504
0.314
0.294
10.699
***
5.432
***
0.168
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
125
Table 7 (cont.)
Components of Earnings and the Persistence of Earnings Within
Accrual Quintiles
Panel B: Operating, Financing and Other Components
Coeff. t-stat Adjusted R
2
Full Sample Intercept
OROE
i,t-1
FROE
i,t-1
Other
i,t-1
0.022
0.565
0.602
0.224
6.126
***
30.408
***
8.635
***
6.782
***
0.251
Quintile 1 Intercept
OROE
i,t-1
FROE
i,t-1
Other
i,t-1
-0.050
0.173
0.035
-0.050
-3.381
***
2.923
***
0.555
-3.381
***
0.024
Quintile 2 Intercept
OROE
i,t-1
FROE
i,t-1
Other
i,t-1
0.003
0.560
0.418
0.003
0.311
6.517
***
3.289
***
0.311
0.231
Quintile 3 Intercept
OROE
i,t-1
FROE
i,t-1
Other
i,t-1
0.021
0.497
0.204
0.021
2.513
**
8.346
***
1.478
2.513
**
0.302
Quintile 4 Intercept
OROE
i,t-1
FROE
i,t-1
Other
i,t-1
0.014
0.616
0.330
0.014
1.595
16.527
***
2.176
**
1.595
0.368
Quintile 5 Intercept
OROE
i,t-1
FROE
i,t-1
Other
i,t-1
0.028
0.509
0.307
0.028
2.609
***
11.209
***
5.842
***
2.609
***
0.163
Panel C: Cash Flow and Accruals Components
Coeff. t-stat Adjusted R
2
Full Sample Intercept
CROE
i,t-1
AROE
i,t-1
0.014
0.545
0.396
3.747
***
30.954
***
20.819
***
0.252
Quintile 1 Intercept
CROE
i,t-1
AROE
i,t-1
-0.050
0.173
0.035
-3.381
***
2.923
***
0.555
0.018
Quintile 2 Intercept
CROE
i,t-1
AROE
i,t-1
0.003
0.560
0.418
0.311
6.517
***
3.289
***
0.164
Quintile 3 Intercept
CROE
i,t-1
AROE
i,t-1
0.021
0.497
0.204
2.513
**
8.346
***
1.478
0.312
Quintile 4 Intercept
CROE
i,t-1
AROE
i,t-1
0.014
0.616
0.330
1.595
16.527
***
2.176
**
0.370
Quintile 5 Intercept
CROE
i,t-1
AROE
i,t-1
0.028
0.509
0.307
2.609
***
11.209
***
5.842
***
0.179
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
126
Table 7 (cont.)
Components of Earnings and the Persistence of Earnings Within
Accrual Quintiles
(3)
All variables as previously defined
*** Two tailed t-test significant at 1% level
** Two tailed t-test significant at 5% level
* Two tailed t-test significant at 10% level
6. Conclusion
This study expands the prior literature
considering the persistence of earnings by
decomposing earnings into components and
determining whether this provides insights into
the persistence of aggregate earnings. Evidence
is provided of differences in persistence
between components of earnings. Differences in
persistence are found between components
formed on the basis of: financial ratios;
operating and financing activities; and cash and
accruals. This supports the decomposition of
earnings into components and their separate
evaluation in financial statement analysis.
However, while there was evidence that this can
be exploited to better explain aggregate earnings
persistence, this result appears to be attributable
primarily to firms with extreme negative
accruals. Accordingly, there is support for the
decomposition of earnings as commonly
advocated in financial statement analysis.
A number of limitations where identified
during the undertaking of this study. First, there
is the relatively small sample size. Where the
relative magnitudes of regression coefficients
prima facie suggest a difference in persistence,
but F-statistics do not reject the equality of
those coefficients, this could simply be a
function of the limited sample. Second, the data
requirements resulted in many firm-year
observations being deleted. This introduced a
survival bias into the sample. Third,
inconsistencies in disclosure requirements and
practices for abnormal, extraordinary and
significant items made the interpretation of the
‘Other’ item variables difficult. Finally, the
usual pattern of multi-collinearity and
autocorrelation issues are inherent in the pooled
autoregression of earnings components.
A number of extensions to the study are
suggested. These include widening the research
and considering alternative decompositions of
earnings. The decompositions considered here
are those most commonly suggested in the
literature. However, it is not exhaustive. Lev
(1973) decomposes the balance sheet into
current and non-current components. Wu &
Fargher (2007) report significantly different
persistence between profits and losses. No
consideration was given in this paper to the
economic or other characteristics of the firm
and how this affects the persistence of earnings.
Introducing controls for these into models of
earnings persistence is suggested. Finally,
consideration was not given to general
economic conditions and how this affects the
persistence of earnings.
References
Cheng, Q., 2005, “What determines residual income?”, The
Accounting Review, vol. 80(1), pp. 85–112.
Dechow, P., 1994, “Accounting earnings and cash flows as
measures of firm performance: The role of accounting
accruals”, Journal of Accounting and Economics,
vol. 18, pp. 3–42.
Dechow, P.M., & W. Ge, 2005, “The persistence of earnings
and cash flows and the role of special items:
Implications for the accrual anomaly”, Review of
Accounting Studies, vol. 11(2-3), p. 253.
Dechow P.M., A.P. Hutton & R.G. Sloan. 1999, “An
empirical assessment of the residual income valuation
model”, Journal of Accounting and Economics,
vol. 26(1-3), pp. 1–34.
Dechow, P.M., S.A. Richardson & R.G. Sloan, 2004, “The
persistence and pricing of the cash component of
earnings”, Working Paper, University of Michigan.
Easton P., & M. Zmijewksi, 1989, “Cross-sectional variation
in the stock market response to accounting earnings
announcements”, Journal of Accounting and
Economics, vol. 11, pp. 117–141.
Elliott J.A., & J.D. Hanna, 1996, “Repeated accounting
write-offs and the information content of earnings”,
Journal of Accounting Research, vol. 34, pp. 135–155
Fairfield P.M., R.J. Sweeney & T.L. Yohn, 1996,
“Accounting classification and the predictive content of
earnings”, The Accounting Review, vol. 71(3),
pp. 337–355.
ROE
it
= ?
0
+ ? ?
j
Component
it-1
+ ?
it
k
j=1
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
127
Fairfield, P.M., S. Whisenant & T. Yohn, 2003, “Accrued
earnings and growth: Implications for future
profitability and market mispricing”, The Accounting
Review, vol. 78(1), pp. 353–371.
Fairfield, P.M., & T.L. Yohn, 2001, “Using asset turnover
and profit margin to forecast changes in profitability”,
Review of Accounting Studies, vol. 6(4), pp. 371–385
Feltham G.A., & J.A. Ohlson, 1995, “Valuation and clean
surplus accounting for operating and financing
activities”, Contemporary Accounting Research,
vol. 11(2), pp. 689–731
Francis J., R. LaFond, P.M. Olsson & K. Schipper, 2004,
“Costs of equity and earnings attributes”, The
Accounting Review, vol. 79(4), pp. 967–1010.
Herrmann D., T. Inoue & W.B. Thomas, 2000, “The
persistence and forecast accuracy of earnings
components in the USA and Japan”, Journal of
International Financial Management and Accounting,
vol. 11(1), pp. 48–70
Palepu, K, P. Healy and V. Bernard, 2004, Business analysis
and valuation: Using financial statements, 3rd ed,
Thomson Southwestern, Ohio.
Kormendi R., & R. C. Lipe. 1987, “Earnings innovations,
earnings persistence, and stock returns.” Journal of
Business, vol. 60(3), pp. 323–345.
Lev, B. 1973, “Decomposition measures for financial
analysis”, Financial Management, vol. 2(1), pp. 56–63.
Modigliani, F., & Miller M.H., 1958, “The cost of capital,
corporation finance and the theory of investment”, The
American Economic Review, vol. 48(3), pp. 261–297.
Nissim D., & S.H. Penman, 2001, “Ratio analysis and equity
valuation: From research to practice”. Review of
Accounting Studies, vol. 6, pp. 109–154.
Ohlson J.A., 1995, “Earnings, book values and dividends in
equity valuation”, Contemporary Accounting Research,
vol. 11(2), pp. 661–687.
Ou, J.A., 1990, “The information content of non earnings
accounting numbers as earnings predictors”, Journal of
Accounting Research, vol. 28(1), pp. 144–163
Penman, S., 2007, Financial statement analysis and security
valuation, 3rd ed, McGraw Hill
Penman S.H. & X.-J. Zhang, 2002, “Modelling sustainable
earnings and P/E ratios with financial statement
analysis”, Working Paper, Columbia University.
Richardson S.A., R.G. Sloan, M.T. Soliman & I. Tuna,
2005, “Accrual reliability, earnings persistence, and
stock prices”, Journal of Accounting and Economics,
vol. 39, pp. 437–485.
Sloan R.G., 1996, “Do stock prices fully reflect information
in accruals and cash flows about future earnings?”, The
Accounting Review, vol. 71(3), pp. 289–315.
Stickney, Brown and Wahlen, 2005, Financial reporting,
financial statement analysis, and valuation. 6th ed.
Thomson South Western.
Watson J., & P.A. Wells, 2005, “The association between
various earnings and cash flow measures of firm
performance and stock returns: some Australian
evidence.” Working Paper, University of Technology,
Sydney.
Wu, S.H & Fargher, N., 2007, “The implications of
accounting distortions, growth and losses for accruals
and profitability: Australian evidence”, Accounting
Research Journal, Forthcoming.
Xie, H. 2001, “The mispricing of abnormal accruals”, The
Accounting Review, vol. 76, pp. 357–373.
Zhang, X.F., 2007, “Accruals, investment, and the accrual
anomaly”, The Accounting Review, forthcoming.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_648851213.pdf
Forecasting future period profitability is widely
identified as an aim of financial statement
analysis, and these forecasts are typically relied
upon for the estimation of firm value. To
facilitate this, the decomposition of earnings
into its components or drivers, is typically
advocated.
Accounting Research Journal
Earnings Decomposition and the Persistence of Earnings
Stephen Kean Peter Wells
Article information:
To cite this document:
Stephen Kean Peter Wells, (2007),"Earnings Decomposition and the Persistence of Earnings", Accounting Research
J ournal, Vol. 20 Iss 2 pp. 111 - 127
Permanent link to this document:
http://dx.doi.org/10.1108/10309610780000696
Downloaded on: 24 January 2016, At: 21:02 (PT)
References: this document contains references to 0 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 477 times since 2007*
Users who downloaded this article also downloaded:
David Heald, Ron Hodges, (2015),"Will “austerity” be a critical juncture in European public sector financial reporting?",
Accounting, Auditing & Accountability J ournal, Vol. 28 Iss 6 pp. 993-1015 http://dx.doi.org/10.1108/AAAJ -04-2014-1661
J oel D. Goldhar, David Lei, (1991),"The Shape of Twenty-first Century Global Manufacturing", J ournal of Business Strategy,
Vol. 12 Iss 2 pp. 37-41 http://dx.doi.org/10.1108/eb039401
Thomas Li-Ping Tang, J wa K. Kim, Theresa Li-Na Tang, (2002),"Endorsement of the money ethic, income, and life
satisfaction: A comparison of full-time employees, part-time employees, and non-employed university students", J ournal of
Managerial Psychology, Vol. 17 Iss 6 pp. 442-467 http://dx.doi.org/10.1108/02683940210439388
Access to this document was granted through an Emerald subscription provided by emerald-srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service
information about how to choose which publication to write for and submission guidelines are available for all. Please visit
www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio of
more than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of online
products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on Publication
Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.
*Related content and download information correct at time of download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
111
Earnings Decomposition and the
Persistence of Earnings
Stephen Kean and Peter Wells
School of Accounting
University of Technology
Abstract
Forecasting future period profitability is widely
identified as an aim of financial statement
analysis, and these forecasts are typically relied
upon for the estimation of firm value. To
facilitate this, the decomposition of earnings
into its components or drivers, is typically
advocated. This paper investigates the existence
of systematic differences in persistence across
the components of earnings. If components of
earnings experience differences in persistence,
this may provide insights into the determinants
of aggregate earnings level and persistence.
This paper provides evidence of differences in
persistence between components of earnings.
Differences are found between components
formed on the basis of: financial ratios;
operating and financing activities; and cash and
accruals. Furthermore, there is evidence that
earnings components improve the explanatory
power of models evaluating aggregate earnings
persistence, with this result being strongest for
firms with extreme income decreasing accruals.
Due to the pivotal role of earnings in firm
valuation, the results from this paper have direct
implications for valuation.
1. Introduction
The objective of this paper is to investigate
the existence of systematic differences in
persistence across the components of earnings,
and to determine whether these can be utilised
to improve the prediction of future period
performance. The bases for the decomposition
Acknowledgements: We appreciate the financial support of
the University of Technology, Sydney. Prior versions of this
paper have benefited from the comments of Alan Hodgson,
Anne Wyatt and participants at the 2006 AFAANZ
Conference and workshops at the University of Technology
Sydney.
Keywords: Financial reporting, earnings persistence.
of earnings evaluated include: component
financial ratios as suggested by financial
statement analysis texts such as Healy, Palepu
and Bernard (2004), Stickney, Brown and
Wahlen, (2005) and Penman (2007); operating
and financing activities (Fairfield, Sweeney and
Yohn, 1996); and cash and accrual components
(Sloan, 1996). Evidence is provided of
significant differences in persistence across
earnings components. There is also evidence of
the earnings components providing insights into
the persistence of earnings, with this result
being strongest for firms with extreme income
decreasing accruals.
When undertaking financial statement
analysis and business valuation, significant
efforts are expended on evaluating the historic
performance of the firm, and this is undertaken
with the aim of informing the forecasting of
future period performance. To facilitate this, the
decomposition of earnings into components is
commonly advocated. A primary motivation for
this paper is the determination whether earnings
decomposition aids the forecasting of future
period performance. Furthermore, if multiple
bases for decomposition are suggested, do
alternative decompositions better satisfy this
function? Earnings are the product of a range of
factors, including the application of accounting
practices (e.g., GAAP), the economic decisions
of the firm and its economic environment
(Francis, LaFond, Olsson & Schipper, 2004).
These factors have potentially distinct impacts
on the level and persistence of aggregate
earnings. Accordingly, analysis of the
persistence of alternative decompositions of
earnings, as suggested by Fairfield and Yohn
(2001), is undertaken.
Based on a sample of 2,845 ASX firm-year
observations over the period 1996-2004, evidence
is provided of significant differences in
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
112
persistence across earnings components,
including earnings decomposed based on:
financial ratios; operating and financing activities;
and cash and accrual components. Furthermore,
there is evidence of earnings components
improving the explanatory power of models
evaluating aggregate earnings persistence,
particularly for partitions of firms with extreme
income decreasing accruals. Such differences
support the decomposition of total earnings into
components for the purpose of persistence
analysis and financial statement analysis.
This paper contributes to the extant literature
as follows. First, earnings are the product of
mechanical accounting practices. Understanding
the operation of these practices, and in particular
how they affect the components of earnings,
provides insights into the persistence of
aggregate earnings. Second, earnings also reflect
management decision making. Understanding
how these manifest in earnings components
earnings also potentially provides insights into the
persistence of aggregate earnings.
The remainder of this paper is organised as
follows. Section 2 provides a review of the
existing literature and develops hypotheses.
Research design and sample selection are
addressed in sections 3 and 4. The results are
presented in section 5, and the conclusions and
suggestions for future research are discussed in
section 6.
2. Theory and Hypothesis
Development
The roles of earnings levels and persistence in
determining firm value are well recognised in
the literature. In analytical papers considering
the residual income model of firm valuation,
information on earnings level and persistence is
captured in either the determination of forecast
future period residual income, or as ‘other
information’ (Feltham & Ohlson, 1995; Ohlson,
1995). Further, there is empirical support for
earnings persistence influencing firm value with
Kormendi & Lipe (1987) and Easton &
Zmijewski (1989) finding more persistent
earnings eliciting more pronounced stock
returns at announcement date. The importance
of earnings persistence is identified in practice,
where it has a significant impact on the
determination of forecast future earnings and
appropriate terminal value assumptions (Nissim
& Penman, 2001). Reflecting this, considerable
attention has been devoted to evaluating
earnings persistence. For example, Dechow,
Hutton & Sloan (1999) considered a number
of model specifications to evaluate the extent
of earnings persistence, while Cheng
(2005) examined a range of ‘value-creating’
(economic) and ‘value-recording’ (accounting)
drivers of aggregate earnings persistence.
However, the persistence of the components
of earnings, and its implications for aggregate
earnings persistence has received relatively
limited attention. Forecasting future earnings
using current earnings involves making the
distinction between the permanent and
transitory nature of earnings. If certain
components of earnings are more permanent or
transitory than other components, then the
corresponding decomposition of earnings will
improve the forecasting of future earnings.
There is evidence of the decomposition of
earnings into profit margin and asset turnover
providing insights into future period earnings
(Fairfield and Yohn, 20001). Additionally, the
differing persistence of cash flows and accruals
has been considered (e.g., Sloan, 1996,
Fairfield, Whisenhant and Yohn, 2003),
although the primary concern here has been
evaluating the ‘accrual anomaly’ rather than the
use of components to explain future period
earnings. Accordingly, the issue addressed here
is whether there are systematic differences in
the persistence of earnings components, and
whether this can be exploited to improve the
prediction of future period earnings.
2.1 The decomposition of earnings into
financial ratio components
Financial statement analysis texts such as
Healy, Palepu and Bernard (2004) and Penman
(2007) typically advocate the decomposition of
earnings into its financial ratio components.
Financial ratios, the “building blocks of
forecasting,” relate financial statement
components to return on equity (Nissim &
Penman, 2001). This identifies the components
of earnings, or more specifically returns as
distinct, but complimentary, ‘drivers’ of future
earnings and generally takes the following
form:
1
1 For detailed decomposition of ROE into these
components refer to Penman (2007).
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
113
( ) [ ] ( )
t t t t t
SPREAD NFL ATO PM ROE * * + =
Where:
ROE
t
: Return on Equity for period t
PM
t
: Profit margin calculated as NOPAT
divided by sales for period t
NOPAT
t
: Net Operating Profit After Tax for
period t
ATO
t
: Asset turnover calculated as Sales
over NOA
ave
for period t
NOA
ave
: Average Net Operating Assets for
period i
NFL
i
: Leverage calculated as NFO
ave
divided by equity
ave
for period t
NFO
ave
: Average Net Financing Obligations
for period t
Equity
ave
: Average equity for period t
SPREAD
t
: Spread calculated as OROA –
NBC for period t
OROA
t
: Return on operating assets
calculated as NOPAT divided by
NOA
ave
for period t
NBC
t
: Net borrowing costs calculated as
NFE divided by NFO
ave
for period t
NFE
t
: Net Financing Expense, net of tax
shelter for period t
A feature of this approach is that it focuses
attention on the determinants of earnings and
captures information about both management
strategies and the economic environment within
which the firm operates.
Evidence on the persistence of the ratio
components of earnings is provided in Nissim
& Penman (2001), and it is inferred that forecast
return accuracy would be improved if earnings
were decomposed in this manner. However, the
portfolio approach adopted in evaluating
persistence limits the ability to demonstrate this
empirically. Some evidence of ratio
decomposition enhancing forecast return
accuracy is provided in Fairfield and Yohn
(2001). They find that decomposing change in
return on assets into change in asset turnover
and change in profit margin is useful in
forecasting next period change in return on
assets. However, there is no evidence of this
improving forecast accuracy on a levels basis.
This suggests further evaluation of the relation
between future period returns and lagged
financial ratio components of earnings.
2.2 The decomposition of earnings into
operating and financing components
A simpler approach to earnings decomposition
would be the classification of earnings by
nature. This approach is adopted in Fairfield
Sweeney and Yohn (1996), who decomposed
earnings into non-recurring, special, operating
and non-operating components. Supporting this
decomposition of earnings, they find
monotonically greater average explanatory
power (measured by adjusted r
2
) with greater
levels of disaggregation. Insights into the
determinants of this result are provided by
Hermann, Inoue & Thomas (2000) who find
significant differences in the persistence of such
components of earnings.
A similar decomposition of earnings is
suggested here, although a distinction between
operating and financing activities consistent
with Nissim & Penman (2001) is followed.
Modigliani & Miller (1958) suggest it is the
relatively uncertain returns from operating
activities of the firm that generate value, while
the financing activities have a zero net present
value. Due to the relatively fixed nature of
interest rates, Hermann, Inoue & Thomas
(2000) suggest that financing income is more
persistent than operating income. Similarly, by
their non-recurring nature, ‘other’ accounting
items
2
will be less persistent (more transitory)
than operating income.
2.3 The decomposition of earnings into cash
and accrual components
Finally, many studies have decomposed
earnings into cash flows and accruals, and
considered the information conveyed by the
separate components. For example, Dechow
(1994) examined the extent to which cash flows
and accruals capture firm performance, finding
that accruals enhanced income numbers as
measures of firm performance. This result was
more pronounced for firms with more volatile
cash flows. Similarly, Sloan (1996) provides
evidence of cash flow components of earnings
being more persistent than accruals, and
contrary to expectation this appears not to be
fully reflected in share prices. Subsequent
studies have further decomposed cash flow and
2 For example, abnormal-items, extraordinary-items, gain-
on-sale of non-current assets, and gains on foreign
currency transactions.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
114
accrual components of earnings (e.g., Xie,
2001; Dechow & Ge, 2005; Richardson, Sloan,
Soliman & Tuna, 2005, Zhang, 2007).
However, these consider the relevance of
earnings components relative to share prices,
rather than evaluating earnings persistence and
whether it benefits forecasting. Sloan (1996)
suggests some accruals are inherently more
transitory than other accruals, and cash flows
are less vulnerable to distortion than accruals.
Fairfield et al (2003) suggest the lower
persistence of accruals reflects accounting
conservatism and the diminishing marginal
return on new project investments. Richardson,
Sloan, Soliman & Tuna (2005) highlight the
subjectivity in the estimation of accruals and a
greater vulnerability to intentional and
unintentional accounting distortions.
Importantly, the tendency for accruals to reverse
is identified, and this is reflected in models of
the persistence of accruals. However,
unresolved is whether separation of earnings
into cash flow and accrual components will
better explain future period earnings.
3. Research Design
3.1 Establishing a benchmark
Central to the evaluation of whether the
decomposition of earnings improves the
evaluation of earnings persistence is the
determination of an appropriate benchmark. To
avoid scale problems earnings are scaled by
equity, and an autoregressive model of earnings
scaled by book value of equity (i.e., ROE) is
utilised. In the literature a number of variations
of book value of equity are used (i.e., opening,
average and closing), and consideration is first
given to the determination of the variation to be
used in this study.
3
Accordingly the following
model is estimated:
(1)
where:
ROE
it
:
it
it
BVEquity
Earnings
3 Research examining earnings persistence often utilises
lagged Equity
i,t-1
(e.g. Ohlson 1995, Fairfield et al 1996,
Dechow et al 1999 Herrmann et al 2000, Penman &
Zhang 2002, Cheng 2005), possibly to avoid the
endogenous effect of Earnings
i,t
upon Equity
i,t
. However,
Sloan (1996), Fairfield & Yohn, (2001) and Nissim &
Penman (2001) utilise average equity to minimise the
effect of substantial changes in equity through the period.
Earnings
it
: Net Profit After Tax Before
Outside Equity Interests for firm i
in year t
BVEquity
it
: Book Value of equity for firm i in
year t, measured as either
i. Opening value
ii. Average Value
iii. Closing Value
The intercept, ?
0
, represents the average level
of ROE independent of lagged ROE, while the
slope coefficient, ?
1
, represents the average
persistence (or reversion toward zero) of ROE.
4
3.2 Persistence of components of earnings
An initial stage in the evaluation of differences
in the persistence of components of earnings is
the estimation of an autoregressive model for
each of the components. This takes the same
general form as above, and is as follows:
(2)
where Component is represented by the
following variables
Financial Ratio Components
PM
it
:
it
it
Sales
NOPAT
ATO
it
:
1 ? it
it
NOA
Sales
NFL
it
:
1
1
?
?
it
it
Equity
NFO
SPREAD
it
:
( )
it it it
NBC ATO PM ? *
Operating, Financing and Other Components
OROE
it
: Earnings from Operating activities
scaled by lagged book value of
equity, for firm i in year t
FROE
it
: Earnings from Financing activities
scaled by lagged book value of
equity, for firm i in year t
OTHER
it
: Other income scaled by lagged
book value of equity, for firm i in
year t
Cash and Accruals Components
CROE
it
: Cash Flow from Operations scaled
by lagged book value of equity, for
firm i in year t
4 Neither equation (1) nor later equations contain RHS
variables other than earnings and its components. This
reflects the concern in this type of study with variation in
persistence of earnings components, rather than variation
in persistence across firms.
ROE
it
= ?
0
+ ?
1
ROE
it-1
+?
it
Component
it
= ?
0
+ ?
1
Component
it-1
+ ?
it
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
115
AROE
it
: Total Accruals scaled by lagged
book value of equity, for firm i in
year t
Differences in the persistence of the
components of earnings will manifest in
differences in model explanatory power, and co-
efficients on the component lagged values.
3.3 Components of earnings and the
persistence of aggregate earnings
To evaluate whether the decomposition of
earnings provides additional information
relevant to the evaluation of earnings
persistence, the following model is estimated
for the different earnings decompositions (i.e.,
financial ratios; operating, financing and other;
and cash and accruals):
(3)
Whether differences in the persistence of
components of earnings can be exploited to
enhance the estimation of future period earnings
will be identified by improvements in model
explanatory power.
4. Data Collection and Sample
Description
4.1 Data collection
Financial data for companies on the ASX from
1996-2004 was collected from the Aspect-
Huntley database. For inclusion, firm-year
observations are required to have all available
data for two consecutive years of persistence
calculation. Observations with earnings-based
ratios beyond +/-1.00 and balance-sheet-based
ratios above 5.00 were excluded to control for
the effect of outliers. Nissim & Penman (2001)
document relatively low persistence for extreme
deciles of ratios, and thus the exclusion of
extreme-ratio observations will bias the
persistence coefficients upwards. Due to
complications in applying the Nissim &
Penman (2001) methodology of removing the
effect of leverage from operating performance
for financial services firms, firms in the
financial services sector are excluded from the
sample.
The sample selection processes, and the
impact of deletions are reported in Table 1. This
provided a final sample of 2,854 firm-year
observations. This highlights the survivorship
bias that is introduced by the requirement for
three years of financial data, and this is
doubtless exacerbated by the exclusion of
observations with extreme ratios. The number
of observations by industry and year was
considered, and no year appears to dominate the
sample, although consistent with Australian
market characteristics the Materials sector does
comprise 23% of the sample.
4.2 Sample description
Descriptive statistics on the variables are
provided in Table 2. It is notable that 22% of
the sample has negative total earnings (ROE).
While this is more than twice the proportion of
losses in Dechow (1994), which was based
on a sample of U.S. firms over the period
1960–1989, it is less than Watson and Wells
(2005). This is likely a consequence of the
sample selection bias discussed above. Poor
performance is not limited to earnings (ROE),
17% of the sample has negative operating
earnings (OROE and PM), and 15% have
negative cash flow from operations (CROE).
The sample mean for ‘Other’ items was also
Table 1
Sample Selection
Firms Firm-years
All firms included in Aspect database, 1996-2004 2,097 14,560
Less: Financial Services firms 420 2,826
Less: Missing data and IPOs 76 529
Less: Foreign currency reporting firms 289 3,214
Less: Observations with at least one extreme ratio* 628 5,146
Final sample size 1996–2004 684 2,845
* Excluded firms: Sales ? 0, Equity ? 0, Operating Assets ? 0, or Net Operating Assets ? 0, -1>Earnings based ratio’s>1,
0?Balance Sheet based ratio>5
ROE
it
= ?
0
+ ??
j
Component
it-1
+ ?
it
k
j=1
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
116
Table 2
Descriptive Statistics
Variable Mean Median Std. Dev. Minimum Maximum Negative
ROE 0.076 0.090 0.180 -0.888 0.936 22%
RNOA 0.08 0.08 0.16 -0.91 1.00 16%
RLOS -0.01 0.00 0.12 -0.84 0.84 44%
PM 0.044 0.050 0.163 -0.983 0.978 17%
ATO 2.036 1.543 1.698 0.010 9.989 0%
NFL 0.348 0.294 0.634 -0.937 7.076 28%
SPREAD 0.031 0.031 0.269 -0.998 0.995 39%
OROE 0.109 0.115 0.168 -0.888 0.975 17%
FROE -0.020 -0.017 0.042 -0.362 0.827 75%
OTHER -0.013 0.000 0.086 -0.892 0.952 41%
CROE 0.169 0.160 0.213 -0.982 1.000 15%
AROE -0.094 -0.084 0.192 -0.984 0.976 76%
Where:
ROE Return-on-Start-of-Period-Equity Spread OROA-NBC
RNOA Return-on-Net-Operating-Assets OROE Operating-Return-on-Equity
RLOS Return-on-Levered-Operating-Spread FROE Financing-Return-on-Equity
PM Profit-Margin OTHER Other-Return-on-Equity
ATO Asset-Turnover CROE Cash-Return-on-Equity
NFL Leverage AROE Accrual-Return-on-Equity
negative, with 41% of the sample observations
being negative which is consistent with the
increasing incidence of firms reporting negative
unusual items. This has previously been
documented in Elliott & Hanna (1996), and is
possibly motivated by the smoothing of ‘usual’
earnings (Francis et al, 2004). This volatility of
earnings, which may be increasing, is likely to
be problematic when evaluating earnings
persistence.
Spearman and Pearson correlations were
calculated (untabulated) and most of the
variables are significantly correlated using both
parametric and non-parametric tests. This broad
correlation amongst the variables suggests a
substantial multicollinearity problem, and is
common in earnings component analysis
(e.g. Nissim & Penman, 2001).
5. Results
5.1 Establishing a benchmark
Central to the evaluation of how earnings
decomposition provides insights into earnings
persistence is the determination of an
appropriate benchmark. Three alternative
specifications of ROE are considered and the
results (untabulated) are not materially different.
As ROE determined by dividing earnings by
beginning of period equity has the highest
t-statistic on ?
1,
and adjusted R
2
(?
1
=0.486,
t-stat=29.077 and adjusted R
2
=0.229) and lacks
an endogeneity issue arising from the inclusion
of current period earning in the measure of
equity, this paper uses beginning of period
equity as the denominator for all scaled
variables.
However, it is notable that the co-efficient on
lagged ROE (?
1
) is materially lower than that
reported by Sloan (1996), only 0.486 compared
to 0.841.
5
Furthermore, the model has relatively
low explanatory power. This suggests much
lower earnings persistence in the present
sample. To gain insights into the magnitude of
the problem, firms were partitioned into
quintiles based on ROE in period t-1, and this
was compared with their quintile ranking in the
subsequent period. The results are presented in
Table 3, and shows considerable volatility in
quintile membership, with as few as 35.5% of
firms maintaining their quintile ranking across
5 Sloan (1996) utilised operating income scaled by total
assets, not ROE.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
117
Table 3
ROE Persistence — Changes in Quintile Ranking from Period t-1 to t
Evaluation of the stability of ROE between periods by considering differences in firm quintile rankings
across periods
Profit 5 8.8% 4.4% 8.3% 22.8% 55.7%
4 6.9% 7.6% 23.9% 41.3% 20.4%
3 10.4% 26.4% 35.5% 18.5% 9.3%
ROE
i,t
2 24.1% 36.5% 21.6% 10.9% 6.9%
Loss 1 49.8% 25.1% 10.7% 6.5% 7.7%
100% 100% 100% 100% 100%
Quintile 1 2 3 4 5
Loss ROE
i,t-1
Profit
Where:
ROEit :
1 ? it
it
BVEquity
Earnings
Earnings
it
: Net Profit After Tax Before Outside Equity Interests for firm I in year t
Table 4
ROE Persistence within Accrual Quintiles
Evaluation of the persistence of ROE for both the full sample of firms (Panel A) and for firms partitioned
on the basis of accruals relative to cash flow (Panel B)
Panel A: Persistence for Full Sample
(n=2,845) Coeff. t-stat Adjusted R
2
Intercept 0.032 10.114
***
0.229
ROE
i,t-1
0.486
29.077
***
Panel B: Persistence Within Accrual Quintiles
Portfolios constructed on the basis of firms ranked by lagged accruals scaled by lagged absolute cash flow
from operations
Coeff. t-stat Adjusted R
2
Quintile 1 Intercept
ROE
i,t-1
-0.020
0.115
-1.936
2.051
**
0.006
Quintile 2 Intercept
ROE
i,t-1
0.019
0.698
2.366
**
10.283
***
0.156
Quintile 3 Intercept
ROE
i,t-1
0.035
0.640
4.720
***
15.647
***
0.300
Quintile 4 Intercept
ROE
i,t-1
0.019
0.641
2.305
**
18.126
***
0.366
Quintile 5 Intercept
ROE
i,t-1
0.006
0.439
0.666
10.250
***
0.155
(1)
All variables as previously defined
*** Two tailed t-test significant at 1% level
** Two tailed t-test significant at 5% level
* Two tailed t-test significant at 10% level
ROE
it
= ?
0
+ ?
1
ROE
it-1
+ ?
it
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
118
the two periods. For example, 32.2% of lagged
quintile 3 firms increased their quintile ranking,
and 8.3% of firms are ranked in quintile 5 in the
subsequent period. Similarly 32.3% of firms
had a lower quintile ranking, and 10.7% of
firms are ranked in quintile 1 in the subsequent
period.
Prior research has identified earnings
persistence as being sensitive to growth (e.g.,
Fairfield, Whisenant and Yohn, 2003), and
large values of negative accruals have been
attributed to unusual items or write-downs
(Elliot and Hanna, 1996). Extreme negative (or
positive) accruals are likely to be transitory.
Accordingly, model 1 is re-estimated with firms
partitioned into quintiles based on lagged
accruals relative to lagged absolute cash flows.
The results are presented in Table 4 and this
identifies considerable variation across the
partitions. For firms experiencing extreme
income decreasing accruals (i.e., quintile 1)
ROE has relatively low persistence (?
1
=0.115, t-
stat=2.051, adjusted r
2
=0.006). While for firms
with income increasing accruals (i.e., quintile
5), there is more persistence in ROE (?
1
=0.439,
t-stat=10.250, adjusted r
2
=0.155). However, the
persistence is much higher for the middle
quintiles, and in particular quintile 4 (?
1
=0.641,
t-stat=18.126, adjusted r
2
=0.366).
6
Evidence in
Watson and Wells (2005) also suggests that
there is a much greater association between
earnings and stock price for these firms.
Accordingly, for firms where reported earnings
better captures underlying firm performance,
earnings persistence is higher. This confirms the
undertaking of subsequent tests on firms
partitioned in this manner.
5.2 Persistence of components of earnings
Attention is directed initially to the evaluation
of the persistence of components of earnings,
and the results of estimating equation 2 are
presented in Table 5. Panel A reports the results
for financial ratio components of earnings. For
the full sample, the co-efficient on profit margin
(PM) is 0.578, while the adjusted r
2
is 0.324.
Notably, there is considerable variation across
firms partitioned on the basis of lagged accruals.
There is much greater persistence in quintiles 3
and 4, measured by the co-efficient on profit
6 For quintile 3 the results of ?
1
=0.640, t-stat=15.647 and
adjusted r
2
=0.300 are similar.
margin (0.693 and 0.676 respectively), and
adjusted r
2
(0.461 and 0.491 respectively) than
in the remaining quintiles (e.g., for quintile 1,
?
1
=0.412, adjusted r
2
=0.166 and quintile 5,
?
1
=0.547, adjusted r
2
=0.218).
This contrasts with the results for asset
turnover (ATO) and leverage (NFL) which
suggest relatively more persistence across all
firms. This is captured in the regressions based
on the full sample of firms, with the co-efficient
on asset turnover (ATO) being 0.797 and the
adjusted r
2
is 0.678, while the co-efficient on
leverage (NFL) is 0.663 and the adjusted r
2
is
0.522. Finally, there is little persistence in the
margin between operating returns and
borrowing costs (SPREAD), with an adjusted r
2
for the full sample of only 0.059, and a
maximum adjusted r
2
of 0.098 for quintile 4.
In combination these suggest considerable
persistence in asset turnover and leverage across
all firms, while profit margin is most persistent
across quintile 3 and 4 firms. In contrast, for the
margin between operating returens and
borrowing costs there is little persistence.
Panel B reports the results for the operating,
financing and other components of earnings.
For the full sample the co-efficient on operating
returns (OROE) is 0.530 and the adjusted r
2
is
0.288. However, there is considerable variation
across the quintiles. For quintiles 2 to 4 there is
greater persistence in operating returns
(OROE), measured by the co-efficient ?
1
(0.733, 0.538 and 0.603 respectively) and
adjusted r
2
(0.25, 0.315 and 0.408 respectively).
However, there is considerable variation across
quintiles 1 and 5 (for example, in quintile 1
?
1
=0.342 and adjusted r
2
=0.081, and in quintile
5 ?
1
=0.465 and adjusted r
2
=0.226). In contrast
there is less variation in the persistence of
financing returns (FROE) across the quintiles
(?
1
=0.326 to 0.623 and adjusted r
2
=0.157 to
0.379), and little persistence in OTHER with a
maximum adjusted r
2
of 0.020 recorded.
Finally the persistence of cash and accrual
components of earnings is presented in Panel C.
There is little variation in the coefficients on
cash returns (CROE) across quintiles 2 to 4
(?
1
=0.514 to 0.532), and the models exhibit
similar explanatory power (adjusted r
2
=0.249 to
0.278). In contrast, for quintile 1 and 5 firms,
the persistence of cash returns (CROE) is much
lower (?
1
=0.322 and 0.368 respectively), and
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
119
the model explanatory power is much reduced
(adjusted r
2
=0.063 and 0.089 respectively).
With respect to accruals, these are subject to
much lower persistence with a maximum
adjusted r
2
of only 0.082. However, consistent
with the results above, there are not major
differences across quintiles 2 to 4 (?
1
=0.290 to
0.368), which contrasts with the results for
quintile 1 and 5 firms (?
1
=0.100 and 0.041
respectively)
In summary, across all the earnings
decompositions there is considerable variation in
the persistence of the components of earnings,
measured by either the co-efficient on the lagged
component or the explanatory power of the
models. Furthermore, these differences are more
pronounced with firms partitioned on the basis of
accruals relative to cash flows. This raises the
question of whether these differences can be
relied upon to better explain earnings persistence.
Table 5
Persistence of Components of Earnings Within Quintiles of
Firms Partitioned on the Basis of Accruals
Evaluation of the persistence of components of earnings for the full sample of firms and for firms
partitioned on the basis of lagged accruals relative to cash flow. This is undertaken for components
determined on the basis of: financial ratios (Panel A); operating, financing and other activities (Panel B);
and cash and accrual components (Panel C).
Panel A: Persistence of Financial Ratio Components
Coeff. t-stat Adjusted R
2
Full Sample Intercept
PM
i,t-1
0.017
0.578
6.673
***
36.915
***
0.324
Quintile 1 Intercept
PM
i,t-1
-0.008
0.412
-1.067
***
10.664
0.166
Quintile 2 Intercept
PM
i,t-1
0.019
0.625
3.713
***
15.897
***
0.307
Quintile 3 Intercept
PM
i,t-1
0.022
0.693
4.883
***
22.053
***
0.461
Quintile 4 Intercept
PM
i,t-1
0.018
0.676
3.741
***
23.452
***
0.491
Quintile 5 Intercept
PM
i,t-1
0.003
0.547
0.466
12.622
***
0.218
Full Sample Intercept
ATO
i,t-1
0.349
0.797
12.913
***
77.421
***
0.678
Quintile 1 Intercept
ATO
i,t-1
0.500
0.740
7.603
***
24.192
***
0.507
Quintile 2 Intercept
ATO
i,t-1
0.086
1.000
1.780
*
47.147
***
0.796
Quintile 3 Intercept
ATO
i,t-1
0.207
0.893
3.888
***
44.510
***
0.777
Quintile 4 Intercept
ATO
i,t-1
0.366
0.779
5.776
***
35.596
***
0.690
Quintile 5 Intercept
ATO
i,t-1
0.406
0.692
6.264
***
32.373
***
0.648
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
120
Table 5 (cont.)
Persistence of Components of Earnings Within Quintiles of
Firms Partitioned on the Basis of Accruals
Panel A: Persistence of Financial Ratio Components of Earnings
Coeff. t-stat Adjusted R
2
Full Sample Intercept
NFL
i,t-1
0.126
0.663
14.449
***
55.725
***
0.522
Quintile 1 Intercept
NFL
i,t-1
0.167
0.716
6.738
***
22.804
***
0.542
Quintile 2 Intercept
NFL
i,t-1
0.077
0.762
4.821
***
33.868
***
0.506
Quintile 3 Intercept
NFL
i,t-1
0.112
0.568
6.340
***
21.380
***
0.445
Quintile 4 Intercept
NFL
i,t-1
0.087
0.723
5.259
***
24.158
***
0.506
Quintile 5 Intercept
NFL
i,t-1
0.163
0.599
7.800
***
25.920
***
0.477
Full Sample Intercept
SPREAD
i,t-1
0.021
0.246
4.337
***
13.358
***
0.059
Quintile 1 Intercept
SPREAD
i,t-1
-0.047
0.080
-3.657
***
2.015
**
0.005
Quintile 2 Intercept
SPREAD
i,t-1
0.018
0.229
1.942
*
4.811
***
0.038
Quintile 3 Intercept
SPREAD
i,t-1
0.038
0.357
3.738
***
7.738
***
0.094
Quintile 4 Intercept
SPREAD
i,t-1
0.038
0.332
3.336
***
7.904
***
0.098
Quintile 5 Intercept
SPREAD
i,t-1
0.024
0.163
1.980
*
3.719
***
0.022
Panel B: Persistence of Operating, Financing and Other Components of Earnings
Coeff. t-stat Adjusted R
2
Full Sample Intercept
OROE
i,t-1
0.046
0.530
14.667
***
33.904
***
0.288
Quintile 1 Intercept
OROE
i,t-1
0.029
0.342
3.905
***
7.137
***
0.081
Quintile 2 Intercept
OROE
i,t-1
0.025
0.733
3.163
***
13.794
***
0.250
Quintile 3 Intercept
OROE
i,t-1
0.064
0.538
9.230
***
16.193
***
0.315
Quintile 4 Intercept
OROE
i,t-1
0.043
0.603
5.909
***
19.811
***
0.408
Quintile 5 Intercept
OROE
i,t-1
0.034
0.465
4.187
***
12.905
***
0.226
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
121
Table 5 (cont.)
Persistence of Components of Earnings Within Quintiles of
Firms Partitioned on the Basis of Accruals
Panel B: Persistence of Operating, Financing and Other Components of Earnings
Coeff. t-stat Adjusted R
2
Full Sample Intercept
FROE
i,t-1
-0.012
0.435
-16.321
***
27.793
***
0.213
Quintile 1 Intercept
FROE
i,t-1
-0.017
0.374
-10.626
***
11.793
***
0.196
Quintile 2 Intercept
FROE
i,t-1
-0.017
0.623
-3.055
**
11.866
***
0.198
Quintile 3 Intercept
FROE
i,t-1
-0.014
0.326
-9.588
***
11.527
***
0.188
Quintile 4 Intercept
FROE
i,t-1
-0.010
0.355
-7.661
***
10.330
***
0.157
Quintile 5 Intercept
FROE
i,t-1
-0.009
0.556
-6.287
***
18.626
***
0.379
Full Sample Intercept
Other
i,t-1
-0.012
0.117
-7.800
***
6.515
***
0.014
Quintile 1 Intercept
Other
i,t-1
-0.023
0.090
-4.016
***
2.421
**
0.008
Quintile 2 Intercept
Other
i,t-1
-0.013
-0.130
-4.464
***
-1.864
*
0.004
Quintile 3 Intercept
Other
i,t-1
-0.005
0.032
-2.227
**
0.555
0.000
Quintile 4 Intercept
Other
i,t-1
-0.009
0.179
-2.994
***
3.509
***
0.020
Quintile 5 Intercept
Other
i,t-1
-0.016
0.143
-4.519
***
3.431
***
0.019
Panel C: Persistence of Cash Flow and Accruals Components of Earnings
Coeff. t-stat Adjusted R
2
Full Sample Intercept
CROE
i,t-1
0.099
0.447
21.833
***
26.800
***
0.201
Quintile 1 Intercept
CROE
i,t-1
0.079
0.322
6.522
***
6.280
***
0.063
Quintile 2 Intercept
CROE
i,t-1
0.075
0.531
6.221
***
14.574
***
0.271
Quintile 3 Intercept
CROE
i,t-1
0.098
0.514
9.141
***
14.838
***
0.278
Quintile 4 Intercept
CROE
i,t-1
0.086
0.532
8.307
***
13.748
***
0.249
Quintile 5 Intercept
CROE
i,t-1
0.114
0.368
12.973
***
7.526
***
0.089
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
122
Table 5 (cont.)
Persistence of Components of Earnings Within Quintiles of
Firms Partitioned on the Basis of Accruals
Panel C: Persistence of Cash Flow and Accruals Components of Earnings
Coeff. t-stat Adjusted R
2
Full Sample Intercept
AROE
i,t-1
-0.082
0.238
-21.662
***
13.287
***
0.058
Quintile 1 Intercept
AROE
i,t-1
-0.133
0.100
-7.712
***
1.834
*
0.004
Quintile 2 Intercept
AROE
i,t-1
-0.074
0.368
-6.364
***
7.208
***
0.082
Quintile 3 Intercept
AROE
i,t-1
-0.071
0.290
-6.523
***
3.650
***
0.021
Quintile 4 Intercept
AROE
i,t-1
-0.057
0.304
-6.504
***
1.924
*
0.005
Quintile 5 Intercept
AROE
i,t-1
-0.046
0.041
-3.725
***
0.730
0.000
(2)
All variables as previously defined
*** Two tailed t-test significant at 1% level
** Two tailed t-test significant at 5% level
* Two tailed t-test significant at 10% level
5.3 Components of earnings and the
persistence of aggregate earnings
Evidence of whether the decomposition of
earnings into components better explains the
persistence of aggregate earnings is presented in
Table 6, and attention is directed in the first
instance towards the decomposition of earnings
into financial ratio components. When lagged
financial ratio components (PM, ATO, FLEV
and SPREAD) are regressed on ROE the
adjusted r
2
declines to 0.157. However, an issue
with this analysis is that it assumes an additive
relation between the components of earnings,
which is not the case for these financial ratios,
and equation 3 would be misspecified.
Addressing this problems, the financial ratio
components are partially aggregated to provide
a linear model (i.e., RNOA = PM *ATO and
RLOS = FLEV*SPREAD) and regressed on
ROE. The adjusted r
2
for this model increases to
0.244, which confirms the original model as
being poorly specified. Importantly, the co-
efficients on the return on net operating assets
(RNOA) and return attributable to leverage
(RLOS) are significantly different and the
increase in adjusted r
2
is significant (Vuong test
z stat = 20.430). Accordingly, there is evidence
of financial ratio components of earnings
exhibiting differing persistence, and evidence of
this being able to be exploited to provide
insights into the persistence of earnings.
Attention now shifts to the decomposition
of earnings operating, financing and other
components. The co-efficients on operating
and financing components of earnings, OROE
t-1
(?
1
= 0.565, t-stat = 30.408) and FROE
t-1
(?
2
= 0.602, t-stat = 8.635) are not significantly
different from each other (F-stat = 0.317).
However, they are significantly different from
the co-efficient on OTHER
t-1
(?
4
= 0.224,
t-stat = 6.782, F-stat = 84.944). Hence, there is
limited of these components of earnings having
differing persistence. Exploiting this limited
difference in persistence, the overall explanatory
power of the model increased, with the adjusted
R
2
increasing from 0.229 to 0.251 (i.e., an
increase of almost 10%), which is statistically
significant (Vuong test z stat = 20.273).
Consequently, there is again limited evidence of
these components of earnings exhibiting
difference persistence, and this being able to be
exploited to better explain earnings persistence.
Component
it
= ?
0
+ ?
1
Component
it-1
+ ?
it
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
123
Table 6
Earnings Components and the Persistence of Earnings
Evaluation of whether the decomposition of earnings into components increases the explanatory power of
models of earnings persistence.
ROE (Benchmark) Financial Ratios A Financial Ratios B
Operating, Financing
and Other
Cash Flow and
Accruals
Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat
Intercept 0.032 10.114
***
0.018 5.450
***
0.023 4.297
***
0.022 6.126
***
0.014 3.747
***
ROE
i,t-1
0.486 29.077
***
RNOA
i,t-1
0.578 30.245
***
PM
i,t-1
0.254 10.868
***
ATO
i,t-1
0.012 6.350***
RLOS
i,t-1
0.441 13.661
***
NFL
i,t-1
0.015 3.220
***
SPREAD
i,t-1
0.179 10.001
***
OROE
i,t-1
0.565 30.408
***
FROE
i,t-1
0.602 8.635
***
Other
i,t-1
0.224 6.782
***
CROE
i,t-1
0.545 30.954
***
AROE
i,t-1
0.396 20.819
***
N 2845 2845 2845 2845 2845
Adjusted R
2
0.229 0.244 0.157 0.251 0.252
F-stat 845.444
***
459.581
***
118.554
***
318.479
***
480.391
***
(3)
All variables as previously defined
*** Two tailed t-test significant at 1% level
** Two tailed t-test significant at 5% level
* Two tailed t-test significant at 10% level
Finally attention is directed to the
decomposition of earnings into cash and accrual
components. The co-efficients on the cash
component of ROE (?
1
= 0.545, t-stat = 30.954)
and the accrual component of ROE
(?
2
= 0.396,
t-stat = 20.819) are significantly different
(F-stat = 89.129). Hence, there is evidence of
these components of earnings having differing
persistence. Further, the overall explanatory
power of the model increased because of this
decomposition of earnings, with the adjusted R
2
increasing from 0.229 to 0.252 (i.e., an increase
of 10%) which is statistically significant
(Vuong test z stat = 20.377).
Accordingly, there is evidence of these
components of earnings exhibiting differences
in persistence, and this being able to be
exploited to better explain earnings persistence.
5.4 Components of earnings and the
persistence of aggregate earnings by quintiles
Whether earnings decomposition provides
insights into earnings persistence may be
influenced by observations where there is little
persistence in earnings, or where the stability of
earnings masks the benefits of decomposition.
Accordingly, tests of whether the decomposition
of earnings into components better explains the
persistence of aggregate earnings is undertaken
separately for quintiles based on lagged accruals
relative to lagged absolute cash flows.
The results for earnings decomposed into
financial ratio components are presented in
Table 7, Panel A.
7
There is considerable
variation in explanatory power of the model
7 Results for regression of lagged PM, ATO, FLEV and
SPREAD on ROE are not reported due to the previously
identified problem with model specification.
ROE
it
= ?
0
+ ? ?
j
Component
it-1
+ ?
it
k
j=1
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
124
across the quintiles, with the adjusted r
2
ranging
from 0.020 to 0.371. Comparison of the results
with the benchmarks in Table 4 suggests that
only for quintile 1 firms are there benefits from
decomposition (i.e., adjusted r
2
increasing from
.006 to 0.020), and this partition of firms is
likely driving the result in Table 6.
The results for earnings decomposed into
operating, financing and other components are
presented in Panel B. Consistent with prior
results this shows there is considerable variation
in explanatory of the model across the quintiles,
with the adjusted r
2
ranging from 0.024 to
0.368. Comparison of the results with the
benchmarks in Table 4 again suggests that there
are few benefits to this decomposition of
earnings for quintile 3, 4 and 5 firms, with only
minimal increases in adjusted r
2
(0.300 v 0.302,
0.366 v 0.368 and 0.155 v 0.163). The benefits
of decomposition appear largest for quintile 1
and 2 firms that are reporting relatively large
negative accruals where the adjusted r
2
increases (0.006 v 0.024 and 0.156 v 0.231),
although the models continue to suffer from
relative low explanatory power.
The results are generally weaker for earnings
decomposed into cash flows and accruals
components, with the results presented in Panel
C. Again, there is considerable variation in
explanatory power across the quintiles with the
adjusted r
2
ranging from 0.018 to 0.370.
However, comparison of these results with the
benchmarks in Table 4 shows that only for
quintile 1 firms are there limited benefits from
decomposition (i.e., adjusted r
2
increasing from
.006 to 0.018).
In summary, from Table 7 it is apparent that
where there is less persistence in ROE,
particularly quintile 1 firms there is some
support of the decomposition of earnings
providing insights into earnings persistence and
this is driving the result in Table 6. However,
for remaining firms the decomposition of
earnings into components provides only limited
insights in earnings persistence that are not
provided by earnings generally.
Table 7
Components of Earnings and the Persistence of Earnings Within
Accrual Quintiles
Evaluation of the decomposition of earnings provides insights into aggregate earnings persistence for the
full sample of firms and for firms partitioned on the basis of lagged accruals relative to cash flow. This is
undertaken for earnings decomposed on the basis of: financial ratios (Panel A); operating and financing and
other activities (Panel B); and cash and accrual components (Panel C).
Panel A: Financial Ratio Components
Coeff. t-stat Adjusted R
2
Full Sample Intercept
RNOA
i,t-1
RLOS
i,t-1
0.018
0.578
0.441
5.450
***
30.245
***
13.661
***
0.244
Quintile 1 Intercept
RNOA
i,t-1
RLOS
i,t-1
-0.026
0.201
0.011
-2.421
**
3.238
***
0.165
0.020
Quintile 2 Intercept
RNOA
i,t-1
RLOS
i,t-1
0.012
0.770
0.533
1.527
***
10.727
***
6.052
***
0.167
Quintile 3 Intercept
RNOA
i,t-1
RLOS
i,t-1
0.034
0.652
0.564
4.665
***
15.757
***
9.462
***
0.303
Quintile 4 Intercept
RNOA
i,t-1
RLOS
i,t-1
0.019
0.649
0.529
2.330
**
18.348
***
9.063
***
0.371
Quintile 5 Intercept
RNOA
i,t-1
RLOS
i,t-1
0.003
0.504
0.314
0.294
10.699
***
5.432
***
0.168
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
125
Table 7 (cont.)
Components of Earnings and the Persistence of Earnings Within
Accrual Quintiles
Panel B: Operating, Financing and Other Components
Coeff. t-stat Adjusted R
2
Full Sample Intercept
OROE
i,t-1
FROE
i,t-1
Other
i,t-1
0.022
0.565
0.602
0.224
6.126
***
30.408
***
8.635
***
6.782
***
0.251
Quintile 1 Intercept
OROE
i,t-1
FROE
i,t-1
Other
i,t-1
-0.050
0.173
0.035
-0.050
-3.381
***
2.923
***
0.555
-3.381
***
0.024
Quintile 2 Intercept
OROE
i,t-1
FROE
i,t-1
Other
i,t-1
0.003
0.560
0.418
0.003
0.311
6.517
***
3.289
***
0.311
0.231
Quintile 3 Intercept
OROE
i,t-1
FROE
i,t-1
Other
i,t-1
0.021
0.497
0.204
0.021
2.513
**
8.346
***
1.478
2.513
**
0.302
Quintile 4 Intercept
OROE
i,t-1
FROE
i,t-1
Other
i,t-1
0.014
0.616
0.330
0.014
1.595
16.527
***
2.176
**
1.595
0.368
Quintile 5 Intercept
OROE
i,t-1
FROE
i,t-1
Other
i,t-1
0.028
0.509
0.307
0.028
2.609
***
11.209
***
5.842
***
2.609
***
0.163
Panel C: Cash Flow and Accruals Components
Coeff. t-stat Adjusted R
2
Full Sample Intercept
CROE
i,t-1
AROE
i,t-1
0.014
0.545
0.396
3.747
***
30.954
***
20.819
***
0.252
Quintile 1 Intercept
CROE
i,t-1
AROE
i,t-1
-0.050
0.173
0.035
-3.381
***
2.923
***
0.555
0.018
Quintile 2 Intercept
CROE
i,t-1
AROE
i,t-1
0.003
0.560
0.418
0.311
6.517
***
3.289
***
0.164
Quintile 3 Intercept
CROE
i,t-1
AROE
i,t-1
0.021
0.497
0.204
2.513
**
8.346
***
1.478
0.312
Quintile 4 Intercept
CROE
i,t-1
AROE
i,t-1
0.014
0.616
0.330
1.595
16.527
***
2.176
**
0.370
Quintile 5 Intercept
CROE
i,t-1
AROE
i,t-1
0.028
0.509
0.307
2.609
***
11.209
***
5.842
***
0.179
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 20 NO 2 (2007)
126
Table 7 (cont.)
Components of Earnings and the Persistence of Earnings Within
Accrual Quintiles
(3)
All variables as previously defined
*** Two tailed t-test significant at 1% level
** Two tailed t-test significant at 5% level
* Two tailed t-test significant at 10% level
6. Conclusion
This study expands the prior literature
considering the persistence of earnings by
decomposing earnings into components and
determining whether this provides insights into
the persistence of aggregate earnings. Evidence
is provided of differences in persistence
between components of earnings. Differences in
persistence are found between components
formed on the basis of: financial ratios;
operating and financing activities; and cash and
accruals. This supports the decomposition of
earnings into components and their separate
evaluation in financial statement analysis.
However, while there was evidence that this can
be exploited to better explain aggregate earnings
persistence, this result appears to be attributable
primarily to firms with extreme negative
accruals. Accordingly, there is support for the
decomposition of earnings as commonly
advocated in financial statement analysis.
A number of limitations where identified
during the undertaking of this study. First, there
is the relatively small sample size. Where the
relative magnitudes of regression coefficients
prima facie suggest a difference in persistence,
but F-statistics do not reject the equality of
those coefficients, this could simply be a
function of the limited sample. Second, the data
requirements resulted in many firm-year
observations being deleted. This introduced a
survival bias into the sample. Third,
inconsistencies in disclosure requirements and
practices for abnormal, extraordinary and
significant items made the interpretation of the
‘Other’ item variables difficult. Finally, the
usual pattern of multi-collinearity and
autocorrelation issues are inherent in the pooled
autoregression of earnings components.
A number of extensions to the study are
suggested. These include widening the research
and considering alternative decompositions of
earnings. The decompositions considered here
are those most commonly suggested in the
literature. However, it is not exhaustive. Lev
(1973) decomposes the balance sheet into
current and non-current components. Wu &
Fargher (2007) report significantly different
persistence between profits and losses. No
consideration was given in this paper to the
economic or other characteristics of the firm
and how this affects the persistence of earnings.
Introducing controls for these into models of
earnings persistence is suggested. Finally,
consideration was not given to general
economic conditions and how this affects the
persistence of earnings.
References
Cheng, Q., 2005, “What determines residual income?”, The
Accounting Review, vol. 80(1), pp. 85–112.
Dechow, P., 1994, “Accounting earnings and cash flows as
measures of firm performance: The role of accounting
accruals”, Journal of Accounting and Economics,
vol. 18, pp. 3–42.
Dechow, P.M., & W. Ge, 2005, “The persistence of earnings
and cash flows and the role of special items:
Implications for the accrual anomaly”, Review of
Accounting Studies, vol. 11(2-3), p. 253.
Dechow P.M., A.P. Hutton & R.G. Sloan. 1999, “An
empirical assessment of the residual income valuation
model”, Journal of Accounting and Economics,
vol. 26(1-3), pp. 1–34.
Dechow, P.M., S.A. Richardson & R.G. Sloan, 2004, “The
persistence and pricing of the cash component of
earnings”, Working Paper, University of Michigan.
Easton P., & M. Zmijewksi, 1989, “Cross-sectional variation
in the stock market response to accounting earnings
announcements”, Journal of Accounting and
Economics, vol. 11, pp. 117–141.
Elliott J.A., & J.D. Hanna, 1996, “Repeated accounting
write-offs and the information content of earnings”,
Journal of Accounting Research, vol. 34, pp. 135–155
Fairfield P.M., R.J. Sweeney & T.L. Yohn, 1996,
“Accounting classification and the predictive content of
earnings”, The Accounting Review, vol. 71(3),
pp. 337–355.
ROE
it
= ?
0
+ ? ?
j
Component
it-1
+ ?
it
k
j=1
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Earnings Decomposition and the Persistence of Earnings
127
Fairfield, P.M., S. Whisenant & T. Yohn, 2003, “Accrued
earnings and growth: Implications for future
profitability and market mispricing”, The Accounting
Review, vol. 78(1), pp. 353–371.
Fairfield, P.M., & T.L. Yohn, 2001, “Using asset turnover
and profit margin to forecast changes in profitability”,
Review of Accounting Studies, vol. 6(4), pp. 371–385
Feltham G.A., & J.A. Ohlson, 1995, “Valuation and clean
surplus accounting for operating and financing
activities”, Contemporary Accounting Research,
vol. 11(2), pp. 689–731
Francis J., R. LaFond, P.M. Olsson & K. Schipper, 2004,
“Costs of equity and earnings attributes”, The
Accounting Review, vol. 79(4), pp. 967–1010.
Herrmann D., T. Inoue & W.B. Thomas, 2000, “The
persistence and forecast accuracy of earnings
components in the USA and Japan”, Journal of
International Financial Management and Accounting,
vol. 11(1), pp. 48–70
Palepu, K, P. Healy and V. Bernard, 2004, Business analysis
and valuation: Using financial statements, 3rd ed,
Thomson Southwestern, Ohio.
Kormendi R., & R. C. Lipe. 1987, “Earnings innovations,
earnings persistence, and stock returns.” Journal of
Business, vol. 60(3), pp. 323–345.
Lev, B. 1973, “Decomposition measures for financial
analysis”, Financial Management, vol. 2(1), pp. 56–63.
Modigliani, F., & Miller M.H., 1958, “The cost of capital,
corporation finance and the theory of investment”, The
American Economic Review, vol. 48(3), pp. 261–297.
Nissim D., & S.H. Penman, 2001, “Ratio analysis and equity
valuation: From research to practice”. Review of
Accounting Studies, vol. 6, pp. 109–154.
Ohlson J.A., 1995, “Earnings, book values and dividends in
equity valuation”, Contemporary Accounting Research,
vol. 11(2), pp. 661–687.
Ou, J.A., 1990, “The information content of non earnings
accounting numbers as earnings predictors”, Journal of
Accounting Research, vol. 28(1), pp. 144–163
Penman, S., 2007, Financial statement analysis and security
valuation, 3rd ed, McGraw Hill
Penman S.H. & X.-J. Zhang, 2002, “Modelling sustainable
earnings and P/E ratios with financial statement
analysis”, Working Paper, Columbia University.
Richardson S.A., R.G. Sloan, M.T. Soliman & I. Tuna,
2005, “Accrual reliability, earnings persistence, and
stock prices”, Journal of Accounting and Economics,
vol. 39, pp. 437–485.
Sloan R.G., 1996, “Do stock prices fully reflect information
in accruals and cash flows about future earnings?”, The
Accounting Review, vol. 71(3), pp. 289–315.
Stickney, Brown and Wahlen, 2005, Financial reporting,
financial statement analysis, and valuation. 6th ed.
Thomson South Western.
Watson J., & P.A. Wells, 2005, “The association between
various earnings and cash flow measures of firm
performance and stock returns: some Australian
evidence.” Working Paper, University of Technology,
Sydney.
Wu, S.H & Fargher, N., 2007, “The implications of
accounting distortions, growth and losses for accruals
and profitability: Australian evidence”, Accounting
Research Journal, Forthcoming.
Xie, H. 2001, “The mispricing of abnormal accruals”, The
Accounting Review, vol. 76, pp. 357–373.
Zhang, X.F., 2007, “Accruals, investment, and the accrual
anomaly”, The Accounting Review, forthcoming.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
0
2
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_648851213.pdf