Documente Academic
Documente Profesional
Documente Cultură
www.emeraldinsight.com/0307-4358.htm
MF
37,4
316
Abstract
Purpose This study aims to test empirically the validity of the accounting valuation model that is
based on earnings and book values for loss-reporting firms under a conservative accounting regime.
Design/methodology/approach The empirical tests are performed by employing returns models
on data derived from non-financial firms listed on the Athens Stock Exchange for the period
1992-2000.
Findings The empirical results suggest that there is no unique concept of income, which is
applicable, for valuation purposes, in all circumstances. Total income may be the appropriate income
concept to use for the valuation of profit reporting firms but not for loss-reporting ones; for loss-reporting
firms, ordinary income appears to be a more useful concept for valuation purposes. Extraordinary
income was also found to be value relevant. Further, different versions of the accounting valuation
model appear to be more relevant for different groups of firms (groups defined in terms of various firm
characteristics, such as: size, growth opportunities and riskiness.
Practical implications The study examines the informational content of the various earnings and
loss items in the income statement and provides conclusions that are useful for standard setters,
accounting policy makers and market participants.
Originality/value It provides further evidence on the value relevance of losses, as opposed to that of
earnings. It coincides with the development of a new project initiated by the International Accounting
Standards Board, i.e. Reporting Comprehensive Income, concerning the content of the income
statement. The analysis is carried in an accounting environment that adopts only historic cost
accounting for the recording and measurement of assets and liabilities, revenues and expenses.
Keywords Stock exchanges, Financial reporting, Accountancy, Greece, Asset valuation
Paper type Research paper
Managerial Finance
Vol. 37 No. 4, 2011
pp. 316-333
q Emerald Group Publishing Limited
0307-4358
DOI 10.1108/03074351111115287
1. Introduction
In this paper, we examine the value relevance of various concepts of accounting losses,
reported in the income statement, under a conservative accounting regime. More
specifically: first, we test whether losses are valued differently than earnings by investors;
and, second, we disaggregate accounting income into its components and we test whether
this disaggregation improves the earnings (loss) returns relation. The loss components
whose value relevance is tested are the components that were identified in both theoretical
and empirical research as playing an important role in stock valuation, i.e. operating income
(OPt)[1], financial income (FINt), income from ordinary activities (ORDt), income from
extraordinary activities (EXT) and discontinued operations (EXTt) and total income (EPSt).
The authors would like to acknowledge Professor Dimitris Ghicas for useful comments.
Loss components
317
MF
37,4
318
where Pt is the market value of the equity at period t, xt is the clean surplus accounting
income, bvt is the book value of equity, b is the ERC and g is the book value response
coefficient (BVRC).
If equation (1) holds then there is no practical benefit of disclosing the components of
clean surplus accounting income separately. In other words, if total clean surplus income is
disaggregated in any two (or more) components (such as x1t and x2t where x1t x2t xt)
and the researcher estimates equation (2) below:
P t a b1 x1t b2 x2t gbvt
X
k
bk
PX kt X 0 LX kt
BV kt21
c
bk
P t21
P t21
P t21
k
Loss components
319
MF
37,4
320
where:
Rt
Pt
dt
Xkt
PXkt
LXkt
BVt2 1 the beginning of fiscal year book value per common share.
bk
bk0
the BVRC.
In model (3) we assume, first, that the ERC sof profit firms is different from that of loss
firms and, second, that the ERCs of the various income components differ among each
other. Formally stated, we well test the validity of the following two hypotheses:
H1. The ERCs of profit-reporting firms is different from the ERC of loss-reporting
firm.
This hypothesis means that the estimated ERCs will differ between profit- and
loss-reporting firms:
H2. The ERCs of the various income components will vary among each other for
both profit- and loss-reporting firms.
This hypothesis means that if total income is disaggregated into its components, in the
earnings book value valuation model, the estimated ERCs will vary among each other.
The ERCs of income components that are of a more permanent nature will be higher from
the ERCs of income components that are of a more transitory nature and, therefore, more
volatile. Similarly, the ERCs of income components that are measured with a higher
degree of conservatism will be higher from the ERCs of income components that are
measured with a lower degree of conservatism[6].
In this paper, we will test the value relevance of the following income components
against each other:
EPSt
ORDt
OPt
FINt
EXTt
TAXt
the income tax per share (as it is calculated under the income tax payable
method)[7].
Loss components
The models that will be actually tested are, therefore, the following:
PEPS t
LEPS t
BV t21
b2
c1
I
P t21
P t21
P t21
PORDt
LORDt
PEXT t
LEXT t
TAX t
BV t21
b2
c1
c2
d
e
II
R t a b1
P t21
P t21
P t21
P t21
P t21
P t21
Rt a b1
R t a b1
e
POP t
LOP t
PFIN t
LFIN t
PEXT t
LEXT t
b2
c1
c2
d1
d2
P t21
P t21
P t21
P t21
P t21
P t21
III
TAX t
BV t21
f
P t21
P t21
where P stand for profit-reporting firms and L stands for loss-reporting firms.
In model (I), we assume that total net income available to common shareholders is the
value relevant variable that is used by the market in the evaluation of stocks. In models
(II) and (III), we relax this assumption and assume that each income component is valued
differently by the market. More specifically:
.
In model (II), we assume that it is not total income that is value relevant. Investors
differentiate between income from ORD (i.e. the income originating from the
permanent activities of the firm) and the income from EXT (which are of a more
transitory nature). We expect that the ERC of the extraordinary income will be less
from that of the ordinary income because the former will be more volatile.
.
In model (III), we disaggregate, further, ordinary income to operating income and
FIN. The reason for this disaggregation is twofold: first, reported FIN under the
cost method under (over)estimates total FIN and thus the estimated ERC of the
FIN will be over (under)estimated; and, second, there are some evidence that FIN is
used by the reporting entities as an income-smoothing device and that it should
be priced differently from operating income (Warfield and Linsmeier, 1992). For
these reasons, no a priori expectations about the sign and the size of the ERCs of
FIN are formed.
4. The data
The data were obtained from the Statistical Department of the Athens Stock Exchange
and cover all non-financial sector firms for which data were available for the period
1992-2000. We identified 79 non-financial firms for which all relevant data were
available for the nine-years period. This gave us a total of 711 reporting firm years.
In Table I, we present the distribution of positive and negative values of the earnings
components for each of the years that this study covers.
From the figures listed on Table I, we notice that there were 118 firm years in which
the earnings per share (EPS) were negative; this number is reduced to 110 for the income
from ORD and 111 for the income from operations (OP), respectively. Most of the
companies in our sample, however, report negative income from EXT as well as FIN.
This latter finding can be attributed to two factors, first, that Greek companies rely
heavily to bank capital as a major source of debt capital, which means that the interest
expense constitutes a major expense item in most income statements; and, second,
as income from investment on securities is recorded only that realized from either
321
MF
37,4
Year
EPSt
ORDt
OPt
FINt
EXTt
322
1992
1993
1994
1995
1996
1997
1998
1999
2000
Total
14
18
12
10
9
12
14
13
16
118
12
15
17
8
7
10
12
12
17
110
8
10
9
9
9
10
16
20
20
111
50
57
49
48
51
49
52
41
39
436
56
59
55
60
48
54
50
62
52
496
Table I.
Distribution of negative
earnings by year
Notes: Variables definition: EPSt the earnings per share available to common shareholders; ORDt
the income from ordinary activities per common share; OPt the income from operations per common
share; FINt the financial income per common share; EXTt the extraordinary income per common
share
EPSt
ORDt
OPt
FINt
EXTt
41
11
5
22
79
43
10
4
22
79
45
10
9
15
79
14
2
4
59
79
3
1
3
72
79
Notes: Variables definition: EPSt the earnings per share available to common shareholders; ORDt
the income from ordinary activities per common share; OPt the income from operations per common
share; FINt the financial income per common share; EXTt the extraordinary income per common
share
Large
35
9.8
OPt
Non-growth
41
15.8
OPt
OPt
More risky Less risky
57
54
16.0
15.0
Growth
70
15.4
Small
76
21.3
Large
158
44.0
FINt
Non-growth
152
58.9
FINt
FINt
More risky Less risky
216
220
60.6
61.7
Growth
284
62.6
Small
278
78.0
Large
197
55.3
EXTt
Non-growth
176
68.2
EXTt
EXTt
More risky Less risky
248
248
69.6
69.6
Growth
320
70.6
Small
299
83.9
Notes: Variables definition: EPSt the earnings per share available to common shareholders; ORDt the income from ordinary activities per common
share; OPt the income from operations per common share; FINt the financial income per common share; EXTt the extraordinary income per
common share
Loss components
323
Table III.
The distribution of losses
by certain firms
characteristics
MF
37,4
324
as small are defined those whose total assets are below the median while as large
are defined those whose total assets are above the median. From the figures listed on
panel A of Table III, we notice that there is a concentration of negative results to those
firms classified as small. The EPS variable, for example, took a negative value for the
27 per cent of the small firms; the respective figure for the large firms was 5.30 per cent.
Growth opportunities are defined in terms of the change in net fixed assets. A positive
change in net fixed assets implies the existence of growth opportunities (here and after
growth firms) while a negative change in net fixed assets implies the absence of growth
opportunities (here and after non-growth firms). From the 711 firm years in our sample,
453 were classified as offering growth opportunities while 258 were classified as
non-offering growth opportunities. From the figures listed on panel B of Table III, we
notice that the percentage of firms that reported negative income items is distributed
almost equally among growth and non-growth firms.
Riskiness is defined in terms of the debt to equity ratio. The firm years in our sample
are split into two equal groups, i.e. to more risky and less risky, according to the median
value of their debt to equity ratio; as less risky are defined those firms whose debt to
equity ratio is below the median while as more risky are defined those whose debt to
equity ratio is above the median. From the figures listed on panel C of Table III, we notice
that for the ORD variable there is a concentration of negative results to the less risky
firms; this is not the case, however, for the other income items which appear to be
distributed evenly among the more risky and the less risky firms.
In Table IV, we present univariate statistics for the variables included in models
(I)-(III).
Rt
PEXTt
LEXTt
PORDt
LORDt
PEPSt
LEPSt
PFINt
LFINt
POPt
LOPt
TAXES
BVt2 1
Table IV.
Descriptive statistics
Minimum
Maximum
Median
Mean
SD
0.000019
0
20.003570
0
20.000812
0
20.000853
0
20.002569
0
20.000577
20.002925
0.000004
0.195820
0.001516
0
0.015796
0
0.012732
0
0.003343
0
0.013679
0
0
0.065232
0.003478
0
20.000002
0.000066
0
0.000054
0
0
20.000005
0.000081
0
20.000034
0.001473
0.009183
0.000014
20.000058
0.000362
0.000021
0.000319
20.000022
0.000077
20.000095
0.000375
20.000015
20.000079
0.003044
0.000661
0.000002
0.000008
0.000035
0.000003
0.000031
0.000003
0.000011
0.000008
0.000033
0.000002
0.000178
0.005145
Notes: n 711; variables definition: Pt the price six months after fiscal year end; dt the dividend
distributed in year t; Rt (Pt dt Pt2 1)/Pt; EPSt the earnings per share available to common
shareholders; PEPSt EPSt if EPSt . 0; 0 otherwise; LEPSt EPSt if EPSt ,0; 0 otherwise; ORDt
the income from ordinary activities per common share; PORDt ORDt if ORDt . 0; 0 otherwise;
LORDt ORDt if ORDt , 0; 0 otherwise; OPt the income from operations per common share; POPt
OPt if OPt .0; 0 otherwise; LOPt OPt if OPt , 0; 0 otherwise; FINt the financial income per
common share; PFINt FINt if FINt . 0; 0 otherwise; LFINt FINt if FINt , 0; 0 otherwise; EXTt
the extraordinary income per common share; PEXTt EXTt if EXTt .0; 0 otherwise; LEXTt EXTt
if EXTt ,0; 0 otherwise; and BVt2 1 the beginning of fiscal year book value per common share; all
variables but Rt were deflated with Pt2 1
From the figures listed on Table IV, we notice that in many cases, the mean values of the
variables deviate substantially from their median values suggesting that the means
have been influenced by outlying observations. We excluded as outliers all those
observations for which the distance between the fitted regression value and the actual
value was greater than three standard deviations.
5. The empirical results
Table V contains univariate parametric correlation coefficients among the different
variables used in this study.
From the figures listed on Table V, we observe that Rt is highly correlated with all
independent variables. Moderate correlation coefficients are also observed among the
various independent variables of our study. These results indicate the existence of
collinearity among certain independent variables. For this reason, the condition index
suggested by Belsley et al. (1980) was calculated for each equation in order to examine
the presence of multicollinearity. The values obtained are very low suggesting that the
absence of multicollinearity. We also correct for heteroscedasticity using Whites (1980)
heteroscedasticity-consistent covariance matrix.
Table VI presents the empirical results of estimating models (I)-(III)[8].
From the figures listed on Table VI for model (I), we notice that the PEPSt variable is
statistically significant and with a positive sign at a 0.01, as expected and the LEPSt
variable is statistically insignificant. This finding is consistent with that obtained by
Collins et al. (1999). It appears that for loss-reporting firms total income available to
common shareholders is value irrelevant; it is only book value that matters in the
valuation process.
In model (II), we relax the assumption that the market values earnings as a composite
number; rather, we assume that each earnings component is valued differently by
investors. From the figures listed on Table VI for model (II), we notice that earnings
(losses) decomposition to ordinary and extraordinary earnings improves the
explanatory power of the valuation model from 0.818 to 0.838. The two earnings
variables for profit reporting firms, i.e. the PORDt and PEXTt variables, have the
expected positive sign and are statistically significant at a 0.01 and a 0.05 levels,
respectively. The LORDt variable is negatively associated with stock returns and
statistically significant at a 0.01, while the LEXTt variable is positively associated
with stock returns and statistically significant at a 0.01. The results obtained from
estimating model (II), compared with those obtained from model (I) suggest that for
loss-reporting firms, the losses from ORD (i.e. the losses from the permanent activities of
the firm) is a superior measure of firms performance than aggregate losses. The positive
sign that was obtained for the LEXTt variable may imply that investors consider it as an
item manageable by the individual firms.
In model (III), we relax the assumption that the market values the ordinary earnings as
a composite whole; rather, we disaggregate the earnings (losses) from ORD to those
originating form the operating activities of the firm (OP) and those originating from the
financing activities of the firm (FIN) and we assume that each earnings component is
valued differently by investors. From the figures listed on Table VI for model (III), we
notice that this decomposition reduces the explanatory power of the valuation model, from
0.818 (for model I) and 0.838 (for model II) to 0.806. In model (III), all the earnings variables
for profit reporting firms behave as expected. For the loss-reporting firms, however,
Loss components
325
Table V.
Correlation coefficients
PORDt
LORDt
PEPSt
LEPSt
PFINt
LFINt
POPt
LOPt
1.000
TAXESt BVt2 1
1.000
20.360 * 1.000
0.228 * 0.098 * *
1.000
20.210 * 0.974 *
0.085 * *
1.000
0.373 * 0.087 * *
0.867 *
0.083 * *
1.000
0.046
0.043
0.028
0.053
0.035
1.000
0.132 * 2.009
0.115 *
0.009
0.081 * *
0.186 *
1.000
20.227 * 0.137 * 2 0.003
0.104 *
0.004
2 0.127 * 2 0.635 *
1.000
20.035
0.042
0.075
0.042
0.072
2 0.011
0.176 *
0.181 * 1.000
0.291 * 2.937 * 2 0.076
20.932 * 20.061
2 0.531 *
0.091 * * 20.797 * 2.069
1.000
20.165 * 0.212 * 2 0.045
0.200 * 20.052
0.193 * * 2 .237 *
0.230 * 2.010 2 0.164 *
LEXTt
dividend distributed in year t; Rt (Pt dt Pt2 1)/Pt; EPSt the earnings per share available to common shareholders; PEPSt EPSt if EPSt . 0; 0
otherwise; LEPSt EPSt if EPSt , 0; 0 otherwise; ORDt the income from ordinary activities per common share; PORDt ORDt if ORDt . 0; 0 otherwise;
LORDt ORDt if ORDt ,0; 0 otherwise; OPt the income from operations per common share; POPt OPt if OPt .0; 0 otherwise; LOPt OPt if OPt , 0;
0 otherwise; FINt the financial income per common share; PFINt FINt if FINt . 0; 0 otherwise; LFINt FINt if FINt , 0; 0 otherwise; EXTt the
extraordinary income per common share; PEXTt EXTt if EXTt . 0; 0 otherwise; LEXTt EXTt if EXTt ,0; 0 otherwise; and BVt2 1 the beginning of
fiscal year book value per common share; all variables but Rt were deflated with Pt2 1
Notes: n 711; significance at: *a 0.01, * *a 0.05 (two-tailed test); variables definition: Pt the price six months after fiscal year end; dt the
1.000
0.210 * *
1.000
20.228 * *
0.072
0.653 * *
0.057
20.110 * * 2 0.418 *
0.666 * *
0.159 *
20.081 * 2 0.035
0.063
0.008
20.037
2 0.092 * *
0.042
0.056
20.013
0.027
20.802 *
0.027
0.287 *
0.030
PEXTt
326
Rt
PEXTt
LEXTt
PORDt
LORDt
PEPSt
LEPSt
PFINt
LFINt
POPt
LOPt
TAXES
BVt2 1
Rt
MF
37,4
Variables
Constant
PEPSt
LEPSt
PORD
LORDt
POPt
LOPt
PFINt
LFINt
PEXTt
LEXTt
TAXES
BV
Adjusted R 2
Wald test
ERCPEPS ERCLEPS
ERCPORD ERCLORD
ERCPOP ERCLOP
ERCPFIN ERCLFIN
ERCPEXT ERCLEXT
ERCPORD ERCPEXT
ERCLORD ERCLEXT
ERCPOP ERCPFIN ERCPEXT
ERCLOP ERCLFIN ERCLEXT
Model I
0.410 (3.963) *
7.862 (2.983) *
2 0.265 (20.880)
0.369 (5.478) *
0.818
Model II
0.378 (3.887) *
8.855 (1.270) *
2 1.233 (23.770) *
3.014 (2.0253) * *
2.684 (2.604) *
2 3.192 (21.795) * * *
0.334 (5.209) *
0.838
Model III
Loss components
0.327 (2.712) *
5.310 (5.340) *
22.347 (2 2.990) *
7.614 (9.205) *
4.439 (3.642) *
4.365 (1.939) * *
20.894 (2 0.542)
23.386 (2 1.735) * * *
0.377 (3.554) *
0.806
327
4.787 * *
1.721 *
8.233 *
15.353 *
9.060 *
1.185 *
0.001
1.200 *
1.786
8.509 *
only the LOPt variable, has the expected negative sign and is significant at a 0.01; the
LFINt variable is positively associated with stock returns and statistically significant at
a 0.01, while the LEXTt variable is statistically insignificant. The positive sign obtained
for the LFINt variable may be attributed to the cost method that is adopted by the Greek
accounting legislation in reporting the income from investments. As it is known, according
to this method, income is recorded only when a dividend distribution occurs and, therefore,
investors may consider the recorded FIN as a sub-estimation of the true FIN of the firm. On
aggregate, however, the valuation model described by equation (III) is not superior to that
described by either equations (I) or (II)[9].
We also performed Wald tests of equality between the estimated ERCs (the results are
shown on the bottom of Table VI). In most cases the null hypotheses (i.e. that the
estimated ERCs are equal among each other) was rejected. These findings give further
support to both of our hypotheses, i.e. that the ERCs of profit-reporting firms differ from
Table VI.
Regression results for the
various accounting
valuation models
MF
37,4
328
those of loss-reporting firms and, secondly, that different income items have different
degrees of value relevance.
In Table VII, we split the sample into two sub-groups, according to the size of the total
assets and we test the validity of our results for small and large firms. The implicit
assumption is that earnings (losses) and their components are more informative for large
than for small firms.
From the evidence in Table VII, we notice that the explanatory power of all three
valuation models is reduced when firms are split into small and large firms. It appears,
however, that when firms are grouped according to size, model (III) performs better than
either model (I) or (II). Comparing the results presented in Table VII with those presented
in Table VI for model (I) we notice that, with the exception of the FIN variable, for small
loss-reporting firms, all other loss variables are value irrelevant. The TAXt variable is
statistically significant and with a negative sign, only in model (III) at a 0.10 for large
firms and at a 0.05 for small firms.
As it is well known from the literature, conservative accounting combined with an
increase in investment will result in a reduction of current profits which, however, are
expected to sustain an increase in future earnings. In Table VIII, we split the sample into
two sub-groups, growth and non-growth firms. We assume that for non-growth firms
earnings (losses) will be less value -relevant than for growth firms.
From the results listed in Table VIII, we notice the explanatory power of models
(I)-(III) for non-growth firms is higher than that for the total sample (reported on Table VI);
model (II), however, exhibits the highest explanatory power. More specifically:
.
In model (I), the LEPSt variable is significant at a 0.05 only for non-growth firms.
.
In model (II), both the LORDt and LEXTt variables are significant at a 0.10
and a 0.05 only for non-growth firms.
.
In model (III), both the LOPt and the LFINt variables are statistically significant
at a 0.01 for growth firms but not for non-growth firms. In the latter sample,
none of the loss variables is statistically significant.
In Table IX, we split the sample into two sub-groups, i.e. more risky and less risky firms.
As more risky firms, we define all those firms whose debt to equity ratio is greater than the
median debt to equity ratio of the firms in our sample; as less risky firms, we define all those
firms whose debt to equity ratio is smaller than the median debt to equity ratio of the firms
in our sample. We expect that earnings will be more informative for the more risky firms.
From the figures listed in Table IX, we notice that the explanatory power of all three
valuation models is reduced when firms are split into two groups according to their
financial risk. For loss-reporting firms that are classified as more risky, none of the
LEPSt, LORDt and LOPt variables is statistically significant. The LFINt variable is
statistically significant for both groups (see model (II)) while the LEXTt variable is
statistically significant at a 0.01, in both models (II) and (III) only for the risky group.
6. Conclusions
In this paper, we tested the validity of the accounting valuation model that is based on
earnings and book values for loss-reporting firms using data from the Athens Stock
Exchange for the period 1992-2000. In the empirical tests, we employed returns models
and corrected for heteroscedasticity using Whites (1980) heteroscedastisity consistent
covariance matrix. The empirical results suggest that:
1.190 (9.389) *
0.693
2.629 *
2.094 (3.442) *
0.606
0.565
Small
0.430 (2.551) *
4.715 (5.855) *
20.025 (20.019)
Model I
0.267 (0.450)
2.887 (0.810)
1.452 (1.075)
Large
3.701 * *
0.475
0.062
3.592 * *
0.909
1.493
0.895
Model III
(20.900)
(1.117)
(0.967)
(20.522)
(2.517) *
(21.003)
(21.843) * * *
(3.221) *
0.158 (0.276)
Large
2 6.401
2.087
3.263
2 3.504
4.284 (1.889) * * *
1.813 (0.869)
4.906
22.400 (20.242) 2 0.690 (20.288) 2 1.478
27.266 (21.002) 2 3.467 (21.485) 2 1.535
2.048 (3.166) *
1.088 (1.034) *
2.126
0.604
0.702
0.614
3.719 (3.404) *
2 0.695 (20.489)
20.003 (20.001)
1.484 (1.133)
Small
0.381 (2.338) *
Model II
0.154 (0.280)
Large
(2.442) *
(21.142)
(2.921) *
(2.797) *
(1.334)
(20.499)
(22.242) * *
(9.663) *
4.206 * *
4.713 * *
3.808 * *
2.705
2 2.678
3.492
3.819
3.050
2 1.037
2 5.436
1.083
0.705
0.411 (2.579) *
Small
Notes: Significance at: *a 0.01, * *a 0.05, and * * *a 0.10 (two-tailed test); variables definition: Pt the price six months after fiscal year end; dt
the dividend distributed in year t; Rt (Pt dt Pt2 1)/Pt; EPSt the earnings per share available to common shareholders; PEPSt EPSt if EPSt . 0; 0
otherwise; LEPSt EPSt if EPSt , 0; 0 otherwise; ORDt the income from ordinary activities per common share; PORDt ORDt if ORDt . 0; 0 otherwise;
LORDt ORDt if ORDt ,0; 0 otherwise; OPt the income from operations per common share; POPt OPt if OPt .0; 0 otherwise; LOPt OPt if OPt , 0;
0 otherwise; FINt the financial income per common share; PFINt FINt if FINt . 0; 0 otherwise; LFINt FINt if FINt , 0; 0 otherwise; EXTt the
extraordinary income per common share; PEXTt EXTt if EXTt . 0; 0 otherwise; LEXTt EXTt if EXTt ,0; 0 otherwise; and BVt2 1 the beginning of
fiscal year book value per common share; all variables but Rt were deflated with Pt2 1
Constant
PEPSt
LEPSt
PORDt
LORDt
POPt
LOPt
PFINt
LFINt
PEXTt
LEXTt
TAXES
BV
Adjusted R 2
Wald test
ERCPEPS ERCLEPS
ERCPORD ERCLORD
ERCPOP ERCLOP
ERCPFIN ERCLFIN
ERCPEXT ERCLEXT
Variables
Loss components
329
Table VII.
Regression results for the
various accounting
valuation models firm
years grouped by size
Table VIII.
Regression results for the
various accounting
valuation models firm
years grouped by growth
opportunities
Non-growth
0.632 (7.387) *
0.898
6.154 *
1.163 (6.245) *
0.780
5.479 *
0.268 (1.553)
0.115 (0.861)
5.917 (7.258) *
7.866 (1.285) *
21.319 (20.576) 21.790 (22.271) * *
Model I
Non-growth
5.557 *
1.671
(2.377) *
(2.226) * *
(1.614) * * *
(6.479) *
3.976 * *
1.231
4.875
3.836
0.587
0.895
3.663 * *
4.012 (1.356)
20.423 (20.155)
23.068 (21.337)
1.070 (5.944) *
0.782
4.777 (4.234) *
9.199 (7.288)
21.747 (20.590) 21.527 (21.749) * * *
0.122 (0.962)
Model II
0.227 (1.477)
Growth
Model III
(4.943) *
(22.347) *
(3.512) *
(2.848) *
(0.814)
(0.731)
(21.753) * * *
(5.508) *
2.278 *
2.036
7.466 *
4.751
29.564
4.867
3.937
2.435
1.558
23.700
1.034
0.794
0.220 (1.581)
Growth
2.473
4.709 * *
3.392 * * *
3.519 (1.622) * * *
20.543 (20.322)
6.944 (7.419) *
2.644 (1.151)
1.169 (1.947) * *
0.130 (0.039)
25.497 (21.590)
0.615 (5.824) *
0.892
0.101 (0.732)
Non-growth
Notes: Significance at: *a 0.01, * *a 0.05, and * * *a 0.10 (two-tailed test); variables definition: Pt the price six months after fiscal year end; dt
the dividend distributed in year t; Rt (Pt dt Pt2 1)/Pt; EPSt the earnings per share available to common shareholders; PEPSt EPSt if EPSt . 0; 0
otherwise; LEPSt EPSt if EPSt , 0; 0 otherwise; ORDt the income from ordinary activities per common share; PORDt ORDt if ORDt . 0; 0 otherwise;
LORDt ORDt if ORDt ,0; 0 otherwise; OPt the income from operations per common share; POPt OPt if OPt .0; 0 otherwise; LOPt OPt if OPt , 0;
0 otherwise; FINt the financial income per common share; PFINt FINt if FINt . 0; 0 otherwise; LFINt FINt if FINt , 0; 0 otherwise; EXTt the
extraordinary income per common share; PEXTt EXTt if EXTt . 0; 0 otherwise; LEXTt EXTt if EXTt ,0; 0 otherwise; and BVt2 1 the beginning of
fiscal year book value per common share; all variables but Rt were deflated with Pt2 1
Constant
PEPSt
LEPSt
PORDt
LORDt
POPt
LOPt
PFINt
LFINt
PEXTt
LEXTt
TAXES
BV
Adjusted R 2
Wald test
ERCPEPS ERCLEPS
ERCPORD ERCLORD
ERCPOP ERCLOP
ERCPFIN ERCLFIN
ERCPEXT ERCLEXT
Growth
330
Variables
MF
37,4
1.026 (8.226) *
0.779
3.860 *
1.136 (5.391) *
0.702
8.890 *
0.215 (0.972)
0.576 (3.360) *
3.826 (3.160) *
5.428 (7.128) *
1.553 (1.547) 2 1.100 (20.588)
Model I
Risky
Less risky
6.006 *
0.806
7.870 *
1.081
5.458
26.253
25.724
0.947
0.735
4.994 (4.001) *
2 1.290 (20.580)
1.956 (1.408)
1.541 (1.553))
Model III
1.508
8.801 *
3.857 * *
(1.387)
(2 1.533)
(1.139)
(1.807) * * *
(1.609) * * *
(2 2.526) *
(2 2.541) *
(4.401) *
0.161 (0.850)
Risky
1.730
23.164
2.312
3.148
(1.065)
3.674 (1.059)
8.044
(2 2.416) *
4.758 (1.490)
25.688
(2 2.090) * * 2 1.314 (20.568) 26.467
(5.045) *
1.028 (8.257) *
0.954
0.775
0.738
0.573 (3.363) *
Less risky
0.181 (0.951)
Model II
Risky
(2.247) * *
(20.792)
(3.924) *
(1.712) * * *
(1.458)
(0.973)
(20.851)
(7.710) *
3.728 * *
0.073
3.555 * * *
4.186
2 2.256
4.829
3.350
5.015
3.776
2 2.520
0.993
0.775
0.582 (3.474) *
Less risky
Notes: Significance at: *a 0.01, * *a 0.05, and * * *a 0.10 (two-tailed test); variables definition: Pt the price six months after fiscal year end; dt
the dividend distributed in year t; Rt (Pt dt Pt2 1)/Pt; EPSt the earnings per share available to common shareholders; PEPSt EPSt if EPSt . 0; 0
otherwise; LEPSt EPSt if EPSt , 0; 0 otherwise; ORDt the income from ordinary activities per common share; PORDt ORDt if ORDt . 0; 0 otherwise;
LORDt ORDt if ORDt ,0; 0 otherwise; OPt the income from operations per common share; POPt OPt if OPt .0; 0 otherwise; LOPt OPt if OPt , 0;
0 otherwise; FINt the financial income per common share; PFINt FINt if FINt . 0; 0 otherwise; LFINt FINt if FINt , 0; 0 otherwise; EXTt the
extraordinary income per common share; PEXTt EXTt if EXTt . 0; 0 otherwise; LEXTt EXTt if EXTt ,0; 0 otherwise; and BVt2 1, the beginning of
fiscal year book value per common share; all variables but Rt were deflated with Pt2 1
Constant
PEPSt
LEPSt
PORDt
LORDt
POP t
LOP t
PFINt
LFINt
PEXTt
LEXTt
TAXES
BV
Adjusted R 2
Wald test
ERCPEPS ERCLEPS
ERCPORD ERCLORD
ERCPOP ERCLOP
ERCPFIN ERCLFIN
ERCPEXT ERCLEXT
Variables
Loss components
331
Table IX.
Regression results for the
various accounting
valuation models firm
years grouped by debt to
equity ratio
MF
37,4
332
.
For loss-reporting firms, the ERC of total losses, ordinary losses and operating
losses is either negative or statistically insignificant.
For the total sample, income from ORD appears to me more value relevant for
loss-reporting firms than either total income or operating income.
Different concepts of the accounting valuation model may be more value relevant
for different groups of firms (groups defined in terms of various firm
characteristics, such as: size, growth opportunities and riskiness).
Extraordinary income is value relevant although its ERC differs from that of the
other income components. Reporting extraordinary income separately in the
income statement, therefore, provides the users of the accounts with information
useful for valuation purposes.
For standard setters, the results of this study suggest that under conservative accounting,
the ERCs of the various income components are not the same. More conservative
accounting practices will result in the estimation of higher ERCs for the respective income
components. Furthermore, reporting different income components in the income statement
is not a waste. The published income statement must be very analytical since different
income concepts are applied in the valuation of firms with different financial characteristics.
Notes
1. This concept of income has been used extensively in empirical research (Penman and Zhang,
2002).
2. The title of the project is Reporting Comprehensive Income. For more information
concerning this project, look at the site of IASB at the following address: www.iasb.org.uk
3. For a review of the definition of conservatism in accounting, see Watts (2003).
4. Greek accounting legislation requires that reporting entities should report only realised
(from transactions) income but at the same time they should anticipate all losses.
5. These models were also used by Martikainen et al. (1997); they are analogous to the models
used by Collins et al. (1999).
6. Assume, for example, that the income that is value relevant is X and that the true ERC is a.
If, because of conservatism in accounting, the reported income equals kX (where 0 , k , 1),
the rational well-informed investor will use, for valuation purposes, X and not kX. The
researcher who will try to estimate the earnings-returns relation using reported income (i.e. kX)
instead of the true income (i.e. X) will obtain an ERC equal to b that will satisfy the following
relation a bk; since 0 , k , 1, for this relation to hold, b should be greater than a. In other
words, the more conservatism is applied in the measurement of a certain income item, the
higher the estimated ERC for that particular income item will be.
7. In Greece, corporate income tax is recorded according to the tax payable method.
8. The results presented in Tables VI-IX were obtained using yearly dummies. Contrary to
other studies (Martikainen et al., 1997), the inclusion of the yearly dummies improved
drastically the explanatory power of all three models.
9. In order to investigate further, the value relevance of the FIN variable we disaggregated it into
income from securities (SGL) and other financial income (OFIN) and we re-estimated model III.
This disaggregation did not improve the explanatory power of the valuation model. We also
carried a Wald test to test the equality between the estimated ERCs of the SGL and the OFIN
variables and the null hypotheses was accepted.
References
Ballas, A. (1999), Valuation implications of exceptional and extraordinary items, British
Accounting Review, Vol. 31 No. 3, pp. 281-95.
Belsley, D., Kuh, E. and Welsch, R. (1980), Regression Diagnostics: Identifying Influential Data
and Sources of Collinearity, Wiley, New York, NY.
Brief, R.P. and Peasnell, K.V. (1996), Clean Surplus: A Link between Accounting and Finance,
Garland, New York, NY.
Collins, D.W., Pincus, M. and Xie, H. (1999), Equity valuation and negative earnings: the role of
book value of equity, The Accounting Review, Vol. 74 No. 1, pp. 29-61.
Fairfield, P.M., Sweeney, R.J. and Yohn, T.L. (1996), Accounting classification and the predictive
content of earnings, The Accounting Review, Vol. 71 No. 3, pp. 337-55.
Giner, B. and Reverte, C. (1998), The value relevance of the earnings decomposition for the
Spanish Stock Market, paper presented in the 21st Annual Congress of the European
Accounting Association, Antwerp.
Hayn, C. (1995), The information content of losses, Journal of Accounting and Economics,
Vol. 20 No. 2, pp. 125-54.
Lipe, R. (1986), The information content of the components of earnings, Journal of Accounting
Research, Vol. 24, pp. 37-64 (Supplement).
Martikainen, T., Kallunki, J.P. and Perttunen, J. (1997), Finish eranings response coefficients: the
information content of losses, European Accounting Review, Vol. 6 No. 1, pp. 69-81.
Ohlson, J.A. (1995), Equity, book values and dividends in equity valuation, Contemporary
Accounting Research, Vol. 11 No. 2, pp. 661-87.
Ohlson, J.A. (1999), On transitory earnings, Review of Accounting Studies, Vol. 4, pp. 145-62.
Ohlson, J.A. and Penman, S.H. (1992), Disaggregated accounting data as explanatory variables
for returns, Journal of Accounting, Auditing and Finance, Vol. 7, Fall, pp. 553-75.
Penman, S.H. and Zhang, X. (2002), Accounting conservatism, the quality of earnings, and stock
returns, The Accounting Review, Vol. 77 No. 2, pp. 237-64.
Stark, A.W. (1997), Linear information dynamics, dividend irrelevance, corporate valuation and
the clean surplus relationship, Accounting & Business Research, Vol. 27 No. 3, pp. 219-28.
Warfield, T.D. and Linsmeier, T.J. (1992), Tax planning, earnings management, and the
differential information content of bank earnings components, Accounting Review, Vol. 67
No. 3, pp. 546-62.
Watts, R.L. (2003), Conservatism in accounting Part I: explanations and implications,
Accounting Horizons, Vol. 17 No. 3, pp. 207-21.
White, H. (1980), A heteroscedastic consistent covariance matrix and a direct test for
heteroscedasticity, Econometrica, Vol. 48, pp. 721-46.
Corresponding author
Dimosthenis Hevas can be contacted at: hevas@aueb.gr
Loss components
333