You are on page 1of 25

Journal of Accounting and Economics 26 (1999) 43—67

Is comprehensive income superior to net income


as a measure of firm performance?!
Dan Dhaliwal!, K.R. Subramanyam""*, Robert Trezevant"
! College of Business and Public Administration, University of Arizona, Tucson, AZ 85721, USA
" Leventhal School of Accounting and Marshall School of Business, University of Southern California,
Los Angeles, CA, 90089-1421, USA

Received 1 October 1997; received in revised form 1 October 1998

Abstract

With the exception of financial firms, we find no evidence that comprehensive income
is more strongly associated with returns/market value or better predicts future cash
flows/income than net income. Moreover, the only component of comprehensive income
that improves the association between income and returns is the marketable securities
adjustment. Our results do not support the claim that comprehensive income is a better
measure of firm performance than net income. Our results also raise questions about the
appropriateness of items included in SFAS 130 comprehensive income as well as the need
for mandating uniform comprehensive income disclosures for all industries. ! 1999
Elsevier Science B.V. All rights reserved.

JEL classification: M41

Keywords: Capital markets; Summary measures of performance; Comprehensive in-


come; SFAS 130

* Corresponding author. Tel.: 1 213 740 5017; fax: 1 213 747 2815; e-mail: krs@almaak.usc.edu
! This paper has substantially benefited from the comments and suggestions of S.P. Kothari (the
editor), Doug Skinner (the referee and discussant), Ashiq Ali, Ray Ball, Mary Barth, Bill Beaver,
Mary Harris, Sanjay Kallapur, Tom Lys, Lillian Mills, Krishna Palepu, Mort Pincus, Katherine
Schipper, Mark Trombley, Terry Warfield, Ross Watts, John Wild, Jerry Zimmerman and the
conference participants at the 1998 Journal of Accounting and Economics Conference. The research
assistance of Marcella Feng and Irina Moroaica is gratefully acknowledged. Trezevant acknow-
ledges the financial support provided by a KPMG Peat Marwick Faculty Fellowship and an Elaine
and Kenneth Leventhal Research Fellowship.

0165-4101/99/$ — see front matter ! 1999 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 5 - 4 1 0 1 ( 9 8 ) 0 0 0 3 3 - 0
44 D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

1. Introduction

1.1. Overview

The Financial Accounting Standards Board has recently issued Statement of


Financial Accounting Standards no. 130 (SFAS 130), Reporting Comprehensive
Income, which mandates that a new summary measure of firm performance,
comprehensive income, be reported in a firm’s primary financial statements.
SFAS 130 does not require additional disclosures. However, it does require that
several items that were reported previously as direct adjustments to equity (i.e.,
as dirty surplus) be reported as adjustments to net income to arrive at compre-
hensive income. In this paper, we use these dirty surplus items to calculate
comprehensive income and then, in an association study context, evaluate the
relative ability of comprehensive income and net income to summarize firm
performance as reflected in stock returns (see Dechow (1994) for a similar
approach).# We also examine which comprehensive income adjustments im-
prove the ability of income to summarize firm performance. Our study does not
address whether the SFAS 130 disclosures convey ‘new’ information that will be
acted upon by market participants. That is, the analysis in this study cannot
address whether reporting dirty surplus items as part of comprehensive income,
rather than as direct adjustments to equity, will affect the way the market
processes accounting information.

1.2. Background: ¹he debate leading up to SFAS 130

SFAS 130 is the culmination of a long-standing debate in the accounting


profession between the ‘all-inclusive’ (or ‘comprehensive income’) and the ‘cur-
rent operating performance’ concepts of reporting income. This debate has been
at the forefront of accounting standard setting from the 1930s to the present.$
Under the all-inclusive concept, the income statement and balance sheet articu-
late completely (i.e., accounting principles are based on clean surplus) and
income includes all revenues, expenses, gains, and losses, whether extraordinary
or otherwise. In contrast, under the current operating performance concept,
extraordinary and nonrecurring revenues, expenses, gains, and losses are ex-
cluded from income.

# Our study is similar in spirit to studies that examine whether summary leverage measures that
include off balance sheet items such as operating leases and pre-SFAS 87 unfunded pension
obligations are better summary measures of firm risk than summary leverage measures based on
balance sheet items alone. Examples include Bowman (1980) and Dhaliwal (1986), who examine the
association between alternate summary leverage measures and systematic equity risk.
$ See Brief and Peasnell (1996) for a chronological collection of papers relating to this debate.
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67 45

Proponents of the all-inclusive concept argue that income measured on


a comprehensive basis measures firm performance better than other summary
income measures because it includes all changes in the net assets of a firm during
a period from nonowner sources. For example, the American Accounting
Association Financial Accounting Standards Committee (1997) (p. 122) states:
“[a]n analyst’s forecast can be used to value a stock only if it is a forecast of
comprehensive income, and a price/earnings ratio only has a precise interpreta-
tion if the earnings are comprehensive2 [F]or reported income to be most
useful in equity price valuation, it must be comprehensive”.
Proponents of the current operating performance concept, on the other hand,
argue that the inclusion of extraordinary and nonrecurring items impairs the
ability of income to reflect the firm’s long-term cash flow prospects. For
example, Black (1993) (p. 5) states that “[i]f we want to maximize the informa-
tion about value in the earnings figure, and minimize the noise, we can choose
accounting rules that make earnings look more like value and less like change in
value. In other words, we can choose rules that minimize transitory components
of earnings, while leaving the permanent components”.
Since the issuance in 1966 of Accounting Principles Board Opinion no. 9,
Reporting the Results of Operations (Accounting Principles Board, 1966), ac-
counting standard-setting bodies in the United States have generally developed
standards using the all-inclusive concept of reporting income. Further, para-
graph 13 of FASB Concepts Statement no. 5 (FASB, 1984) indicates that income
measured on an all-inclusive basis (i.e., comprehensive income) should be re-
ported in a full set of financial statements and comprehensive income is the only
summary income measure that is included in the ten elements of financial
statements as defined in FASB Concepts Statement no. 6 (FASB, 1985). Despite
the Financial Accounting Standards Board’s acceptance of the all-inclusive
concept of reporting income, various pre-SFAS 130 Statements of Financial
Accounting Standards violated this concept by permitting items of comprehens-
ive income to bypass the income statement and be reported as direct adjust-
ments to equity.%
During the 1990s, the Financial Accounting Standards Board and some
influential user groups (e.g., the Association for Investment Management and
Research, 1993) became increasingly concerned about the erosion of the all-
inclusive concept of reporting income (see paragraphs 36 to 50 of SFAS 130;
Johnson et al., 1995; Beresford et al., 1996; Foster and Hall, 1996; Johnson and

% These items came into existence because, when the Statements of Financial Accounting Stan-
dards to which they related were being deliberated, it was ‘‘not clear that the requisite majority of
Board members would have voted for the adoption of those Statements without the compromises
afforded by bypassing the income statement and taking certain items directly to equity” (Johnson et
al., 1995, p. 131).
46 D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

Swieringa, 1996). This concern was exacerbated by the fact that the Financial
Accounting Standards Board’s financial instruments project, which resulted in
SFAS no. 133, Accounting for Derivative Instruments and Hedging Activities
(FASB, 1998), would create additional items of comprehensive income that
would likely be treated as dirty surplus if a standard for comprehensive income
was not issued.
In response to the preceding concerns, as well as to proposals by the Interna-
tional Accounting Standards Committee for reporting comprehensive income
(IASC, 1996), the Financial Accounting Standards Board issued Statement of
Financial Accounting Standards no. 130, Reporting Comprehensive Income, in
June 1997 (FASB, 1997). SFAS 130 mandates that, for fiscal years beginning
after 15 December 1997, comprehensive income and its component adjustments,
collectively referred to as ‘other comprehensive income’, be reported in a firm’s
primary financial statements. Under SFAS 130, the three items included in
‘other comprehensive income’ are the change in the balance of unrealized gains
and losses on marketable securities, the change in the cumulative foreign
currency translation adjustment, and the change in additional minimum pen-
sion liability in excess of unrecognized prior service costs. Exhibit 1 displays
a stylized financial statement disclosure of SFAS 130 comprehensive income.

1.3. Objectives of the study

Given the previously-described debate between the all-inclusive and the


current operating performance concepts of reporting income, surprisingly there
has been little empirical research examining the claim that income measured on
a comprehensive basis is a better measure of firm performance than other
summary income measures. A major objective of this study is to investigate the
above claim by testing (in an association study context) whether comprehensive
income or net income better summarizes firm performance as reflected in stock
returns.&

& Examining the relative superiority of alternate income measures is important because income is
used as a summary measure of firm performance in many contracting and valuation contexts. For
example, Dechow et al. (1998) (p. 1) note that “[e]arnings occupy a central position in accounting. It
is accounting’s summary measure of a firm’s performance”. Further, Dechow (1994) (p. 4) observes
that “[e]arnings are important since they are used as a summary measure of firm performance by
a wide range of users. For example, they are used in executive compensation plans, in debt
covenants, in the prospectuses of firms seeking to go public, and by investors and creditors’. Also,
Black (1993) (p. 3) notes that users of financial statements are “especially interested in summary
figures like earnings and book value. Summary figures are useful because they convey a lot of
information without requiring much of the user, and because they can incorporate details that the
firm chooses not to disclose separately”.
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67 47

A secondary objective of this study is to examine which components of SFAS


130 ‘other comprehensive income’ improve income’s ability to summarize firm
performance. This analysis allows us to draw inferences regarding the appro-
priateness of current and potential items of comprehensive income. These
inferences should assist the Financial Accounting Standards Board as it turns to
the broader-scope projects (described in SFAS 130, paragraph 54) that will
address the issue of which items should be included in ‘other comprehensive
income’.
To evaluate which components of SFAS 130 ‘other comprehensive income’
improve income’s ability to summarize firm performance, we adjust net income
individually for each component of SFAS 130 ‘other comprehensive income’ and
then compare the association between each of the resulting measures of adjusted
income and returns with the association between net income and returns. These
tests are conducted both for the entire sample and within major industry groups.

1.4. Summary of results

We find no clear evidence that comprehensive income is more strongly


associated (in terms of explanatory power) with returns than net income. We
also find that net income is more strongly associated with the market value of
equity and predicts future operating cash flows and income better than compre-
hensive income. These results do not support the claim that income measured
on a comprehensive basis is a better measure of firm performance than other
summary income measures.
We also find that the only component of SFAS 130 ‘other comprehensive
income’ that improves the association between income and returns is the
marketable securities adjustment. Additional analysis indicates that this finding
is driven by financial sector firms.' An implication of these results is that, with
the exception of the marketable securities adjustment, the components of SFAS
130 ‘other comprehensive income’ merely add noise to comprehensive income.
Also, the insignificant results for non-financial industries raise questions about
the usefulness of mandating uniform comprehensive income disclosures for all
industries. Finally, as discussed more fully in Section 3.5, these results suggest
that to improve the ability of comprehensive income to summarize firm perfor-
mance, the Financial Accounting Standards Board should focus on items that
are closely related to a firm’s operations.

' With the exception of the marketable securities adjustment for financial sector firms, the
component adjustments used to convert net income to comprehensive income are, on average, of
insignificant magnitude. Therefore, it is possible that these results may simply arise from a lack of
materiality. As reported in Section 3.4, we test for this possibility by limiting our analysis to the
respective deciles with the largest component adjustments. The results of this analysis suggest that
a lack of materiality is an unlikely explanation for our full sample results.
48 D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

There is some previous evidence related to the components of SFAS 130


‘other comprehensive income’. However, this evidence is mixed. For example,
Barth (1994), Ahmed and Takeda (1995), Barth et al. (1995), Barth et al. (1996),
Eccher et al. (1996), and Nelson (1996) provide mixed evidence on the associ-
ation between the marketable securities adjustment and returns for banks
and/or thrifts. Also, Bartov (1997) finds some evidence that the level of the
cumulative foreign currency translation adjustment is associated with returns,
although this result is sensitive to the returns window used. Last, O’Hanlon and
Pope (1997) (p. 13) find “little evidence that U.K. dirty surplus accounting flows
contain value-relevant items2”. Our study extends prior research by examin-
ing all three components of ‘other comprehensive income’ together in the
context of evaluating the effectiveness of comprehensive income as a summary
measure of firm performance. Also, this is the first study to examine the
association between the marketable securities adjustment and returns for firms
outside the banking and thrifts industry.
SFAS 130 disclosures are not available until financial statements for fiscal
years beginning after 15 December, 1997 are issued. For this reason, our study
cannot examine whether reporting dirty surplus items as part of comprehensive
income rather than as direct adjustments to equity will affect the way that the
market processes accounting information (Bernard and Schipper, 1994). In
other words, our study cannot determine if the SFAS 130 disclosures convey
‘new’ information to the market that can be expected to cause a change in stock
prices, nor are we suggesting that they should. Rather, our study is an associ-
ation study investigating “whether accounting earnings measurements are con-
sistent with the underlying events and information set reflected in stock prices.
Typically, causality is not inferred. Rather, the focus is on whether the earnings
determination process captures in a meaningful and timely fashion the valuation
relevant events” (Collins and Kothari, 1989) (p. 144).
The study proceeds as follows. We discuss test design issues in Section 2.
Section 3 describes our tests of the association of net income and comprehensive
income and its components with returns. Section 4 reports the results of tests of
the association of summary income measures with price and with future operat-
ing cash flows and income. The paper concludes with a summary of its major
findings in Section 5.

2. Design issues

2.1. Measurement of comprehensive income

SFAS 130 defines comprehensive income as net income adjusted for items of
‘other comprehensive income’. Under SFAS 130, the three items included in
‘other comprehensive income’ are the change in the balance of unrealized gains
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67 49

and losses on marketable securities (MKT-ADJ), the change in the cumulative


foreign currency translation adjustment (FC-ADJ), and the change in additional
minimum pension liability in excess of unrecognized prior service costs (PENS-
ADJ). Prior to the adoption of SFAS 130, a firm would report these three items
on its balance sheet as direct adjustments to equity (i.e., as dirty surplus). To
provide evidence on comprehensive income as it is defined by SFAS 130, we
calculate as-if SFAS 130 comprehensive income as net income adjusted for these
three dirty surplus items. We refer to this measure of as-if SFAS 130 compre-
hensive income as COMP .
!$)
In addition to COMP , we use a broader measure of comprehensive income
!$)
in our tests. This measure is based on the annual change in a firm’s comprehens-
ive retained earnings (COMPUSTAT item !36) which, in addition to accumu-
lated earnings, includes the three dirty surplus items required to calculate
COMP as well as several non-SFAS 130 items. Many of these non-SFAS 130
!$)
items were initially considered for inclusion in ‘other comprehensive income’
(SFAS 130, paragraphs 109 to 112). Also, the Financial Accounting Standards
Board is likely to revisit the issue of the inclusion of these items in ‘other
comprehensive income’ in the future (SFAS 130, paragraph 119). Given these
considerations, we calculate a measure of broadly defined comprehensive in-
come, COMP , as the change in COMPºS¹A¹ item !36 after the effects of
"#$!%
common stock dividends are removed.
SFAS 130 refers to the net effect of the adjustments made to net income to
arrive at comprehensive income as ‘other comprehensive income’. Similarly, we
refer to the difference between COMP (COMP ) and net income as
"#$!% !$)
OTH (OTH ). Our evaluation of the components of OTH requires
"#$!% !$) !$)
that we adjust net income individually for MKT-ADJ (FC-ADJ; PENS-ADJ).
We refer to the resulting measures of income as COMP (COMP ;
*+,&-./ 01&-./
COMP ).
2345&-./

2.2. Sample and descriptive statistics

The sample used to estimate returns models consists of all 1994 and 1995
firm-years that have COMPUSTAT and CRSP data needed to calculate re-
turns, net income, COMP , and COMP .( The sample is limited to the
"#$!% !$)
years 1994 and 1995 because the data required to calculate MKT-ADJ are not
available until 1994. To avoid problems with outliers, firm-years for which the
absolute value of the test variable related to net income (COMP ; COMP )
"#$!% !$)

( We set missing COMPUSTAT data for dirty surplus items equal to zero. We obtain qualitat-
ively similar results if we estimate models using only firm-years that (i) report no missing COMPUS-
TAT data for dirty surplus items; or (ii) report non-zero COMPUSTAT data for at least one dirty
surplus item.
50 D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

falls in the top percentile of the distribution for that test variable are eliminated
from the sample.
The final sample consists of 11,425 firm-years. Later, we report the results of
tests that use the market value of equity (future operating cash flows and
income) as the dependent variable. In these tests, sample size is 11,348 (8,893)
firm-years. The smaller number of firm-years in these tests is primarily the result
of missing COMPUSTAT data.
Table 1 reports descriptive data for the sample used to estimate returns
models. For COMP (COMP ), more than 47% (71%) of sample firm-
!$) "#$!%
years exhibit comprehensive income that differs from net income. Thus, compre-
hensive income adjustments are relatively common in our sample. Moreover,
the mean absolute difference between COMP (COMP ) and net income is
!$) "#$!%
approximately 10% (21%) of the mean absolute value of net income, which
translates to approximately 1.2% (2.6%) of the market value of equity. These
differences are economically large in spite of the fact that, for COMP
!$)
(COMP ), more than 52% (28%) of sample observations are zero and
"#$!%
outliers have been eliminated. Finally, regarding the individual components of
‘other comprehensive income’, the mean absolute values of FC-ADJ and
PENS-ADJ are much smaller than the mean absolute value of MKT-ADJ.

3. Tests of association with returns

3.1. Tests of the association of alternate income measures with returns

To investigate the claim that income measured on a comprehensive basis is


a better measure of firm performance than other summary income measures, we
test whether comprehensive income or net income better summarizes firm
performance as reflected in stock returns using the following models:6
R "! #" *NI ## , (1a)
! ) ! ! !
R "! #" *COMP ## , (1b)
! ) ! "#$!%"! !
R "! #" *COMP ## , (1c)
! ) ! !$)"! !
where (COMPUSTAT item numbers are in parentheses). R is raw daily per-
centage returns compounded over the fiscal year; NI is net income after extra-
ordinary items and discontinued operations (!172); COMP is change in
"#$!%

6 As noted in Subramanyam (1996) (p. 258), the use of earnings levels as a proxy for unexpected
earnings in a regression of returns and earnings has theoretical and empirical support (e.g., Ohlson
and Shroff, 1992; Kothari, 1992; Easton and Harris, 1991; Ali and Zarowin, 1991; 1992). Tests that
use first differences in earnings as a proxy for unexpected earnings yield qualitatively similar results.
Table 1
Descriptive data for variables used to estimate models of the association of net income and comprehensive income with returns

99th 1st Percent non- Mean absolute


Variable! Mean Std. dev. Median percentile percentile zero values value

R 0.147 0.693 0.056 2.837 !0.822 99.5% 0.402


NI 0.009 0.204 0.056 0.351 !0.850 99.9% 0.123
COMP 0.007 0.217 0.049 0.495 !0.889 99.9% 0.129
"#$!%
COMP 0.011 0.209 0.054 0.449 !0.854 99.9% 0.127
!$)
OTH !0.002 0.085 0.000 0.257 !0.212 71.4% 0.026
"#$!%
OTH 0.003 0.047 0.000 0.153 !0.117 47.8% 0.012
!$)
MKT-ADJ 0.002 0.045 0.000 0.147 !0.107 22.7% 0.010
FC-ADJ 0.000 0.009 0.000 0.026 !0.018 26.4% 0.002
PENS-ADJ !0.000 0.010 0.000 0.005 !0.006 8.2% 0.001
COMP 0.011 0.209 0.053 0.450 !0.853 100.0% 0.127
*+,&-./
COMP 0.009 0.203 0.056 0.351 !0.844 99.9% 0.123
01&-./
COMP 0.009 0.204 0.056 0.353 !0.854 99.9% 0.123
2345&-./
Pearson correlation between net income and measures of comprehensive income

COMP COMP COMP COMP COMP


"#$!% !$) *+,&-./ 01&-./ 2345&-./
NI 0.920 0.975 0.974 0.999 0.999

Notes: The sample consists of all 1994 and 1995 firm-years that have COMPUSTAT and CRSP data needed to calculate return, net income, and
comprehensive income. Observations for which the absolute value of the test variable related to net income or comprehensive income falls in the top
percentile of the test-variable distribution are eliminated from the sample. Sample size is 11,425 firm-years.
!R is raw daily percentage returns compounded over the fiscal year; NI is net income after extraordinary items and discontinued operations; COMP is
"#$!%
change in COMPUSTAT item !36, adjusted retained earnings, plus common stock dividends; COMP is as-if SFAS 130 comprehensive income;
!$)
OTH is ‘other comprehensive income’ equal to COMP minus NI; OTH is ‘other comprehensive income’ equal to COMP minus NI;
"#$!% "#$!% !$) !$)
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

MKT-ADJ is change in the balance of unrealized gains and losses on marketable securities; FC-ADJ is change in cumulative foreign currency translation
adjustment; PENS-ADJ is change in additional minimum pension liability in excess of unrecognized prior service cost; COMP is net income
*+,&-./
adjusted for MKT-ADJ; COMP is net income adjusted for FC-ADJ; COMP is net income adjusted for PENS-ADJ. Missing COMPUSTAT
01&-./ 2345&-./
data for MKT-ADJ, FC-ADJ, and/or PENS-ADJ, are set equal to zero. All variables except R are deflated by market value of common equity at the
51

previous fiscal year-end.


52 D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

comprehensive retained earnings plus common stock dividends ($—[!36]


# !21); COMP is net income adjusted for the components of SFAS 130
!$)
‘other comprehensive income’ (!172# $!238# $!230#0.65 times
$[!297 !!298, if less than zero]);7 NI, COMP , and COMP are
"#$!% !$)
deflated by market value of common equity at the previous fiscal year-end.
The results of estimating models 1(a) to 1(c), reported in panel A of Table 2,
are ambiguous. The adjusted R# using COMP is smaller than the adjusted
"#$!%
R# using net income. However, the Vuong (1989) test indicates that the differ-
ence in adjusted R# is not significant at conventional levels. On the other hand,
the adjusted R# using COMP is larger than the adjusted R# using net income
!$)
and the difference in adjusted R#, although small in economic terms (i.e., it
increases from 3.81% to 4.20%), is significant at the 0.01 two-tailed level. The
conflicting nature of these results offers no clear evidence on whether compre-
hensive income or net income better summarizes firm performance. This finding,
in conjunction with the finding in Section 4 that net income is more strongly
associated with the market value of equity and predicts future operating cash
flows and income better than comprehensive income, does not support the claim
that income measured on a comprehensive basis is a better measure of firm
performance than other summary income measures.

3.2. Tests of the components of SFAS 130 ‘other comprehensive income’

The results reported in Section 3.1 suggest that what is included in compre-
hensive income influences its ability to summarize firm performance as reflected
in stock returns (e.g., the association of returns with COMP is smaller than
"#$!%
that of returns with COMP ). To allow us to draw inferences regarding the
!$)
appropriateness of current and potential items of comprehensive income, it is
thus important to examine which components of SFAS 130 ‘other comprehens-
ive income’ improve income’s ability to summarize firm performance. To pro-
vide evidence on this question, we estimate the following models:
R "! #" *COMP — ## , (2a)
! ) ! *+, -./"! !
R "! #" *COMP — ## , (2b)
! ) ! 01 -./"! !
R "! #" *COMP ## , (2c)
! ) ! 23458-./"! !

7 Regarding the accuracy of COMPUSTAT items used to calculate COMP , in a validation


!$)
sample of the annual reports of 345 sample firms, the correlation between $!238 ($!230) and
MKT-ADJ (FC-ADJ) calculated using dirty surplus accounts in annual reports is 0.99 (0.97). The
calculation 0.65 times $(!297!!298, if less than zero) mimics GAAP rules for computing
PENS-ADJ. In the validation sample, the correlation between this COMPUSTAT-based amount
and the amount calculated using dirty surplus accounts in annual reports is approximately 0.40.
Given this measurement error, we repeat our tests after calculating COMP with no adjustment
!$)
for minimum pension liability, with qualitatively similar results.
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67 53

Table 2
Results of the estimation of models that test the association of net income and comprehensive
income with returns: Firms are not classified by industry

Panel A: Association of net income and comprehensive income with returns

Model! INT" NI" COMP " COMP " Adj. R#"'


"#$!% !$)
(1a) 0.141 0.665 3.81
(22.12)*% (21.30)*
(1b) 0.142 0.599 3.51
(22.34)* (20.40)* (!1.19)
(1c) 0.139 0.680 4.20
(21.84)* (22.41)* (4.74)*(

Panel B: Association of net income and component-based comprehensive income with returns

Model! INT" COMP " COMP " COMP " Adj. R#"'
*+,&-./ 01&-./ 2345&-./
(2a) 0.139 0.681 4.22
(21.89)* (22.46)* (6.23)*
(2b) 0.140 0.665 3.81
(22.06)* (21.29)* (0.02)
(2c) 0.141 0.663 3.79
(22.13)* (21.25)* (!1.18)

Notes: The sample consists of all 1994 and 1995 firm-years that have COMPUSTAT and CRSP data
needed to calculate return, net income, and comprehensive income. Observations for which the
absolute value of the test variable related to net income or comprehensive income falls in the top
percentile of the test-variable distribution are eliminated from the sample. Sample size is 11,425.
!Model (1a): R "! #" *NI ##
! ) ! ! !
Model (1b): R "! #" *COMP ##
! ) ! "#$!%"! !
Model (1c): R "! #" *COMP ##
! ) ! !$)"! !
Model (2a): R "! #" *COMP ##
! ) ! *+,&-./"! !
Model (2b): R "! #" *COMP ##
! ) ! 01&-./"! !
Model (2c): R "! #" *COMP ##
! ) ! 2345&-./"! !
Model variables are defined as follows: R is raw daily percentage returns compounded over the fiscal
year; NI is net income after extraordinary items and discontinued operations; COMP is change
"#$!%
in COMPUSTAT item !36, adjusted retained earnings, plus common stock dividends; COMP
!$)
is as-if SFAS 130 comprehensive income; OTH is ‘other comprehensive income’ equal to
"#$!%
COMP minus NI; OTH is ‘other comprehensive income’ equal to COMP minus NI;
"#$!% !$) !$)
MKT-ADJ is change in the balance of unrealized gains and losses on marketable securities; FC-ADJ
is change in cumulative foreign currency translation adjustment; PENS-ADJ is change in additional
minimum pension liability in excess of unrecognized prior service cost; COMP is net income
*+,&-./
adjusted for MKT-ADJ; COMP is net income adjusted for FC-ADJ; COMP is net
01&-./ 2345&-./
income adjusted for PENS-ADJ. Missing COMPUSTAT data for MKT-ADJ, FC-ADJ, and/or
PENS-ADJ, are set equal to zero. All variables except R are deflated by market value of common
equity at the previous fiscal year-end.
54 D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

Table 2 (Notes continued)

"The columns labeled ‘INT’, ‘NI’, ‘COMP ’, ‘COMP ’, ‘COMP ’, ‘COMP ’, and
"#$!% !$) *+,&-./ 01&-./
‘COMP ’ report the estimated coefficient on the relevant model variable and, in parentheses,
2345&-./
its associated t-statistic.
'Adjusted R# are compared using the likelihood ratio test described in Vuong (1989). The number in
parentheses is the z-statistic associated with the Vuong test.
%* in the columns identified in note b indicates that an estimated coefficient is greater than zero at the
0.01 one-tailed significance level.
(* in the Adj. R# column indicates that adjusted R# in the estimate of model (1a) differs from adjusted
R# in the estimate of model (1b),2,(2c) at the 0.01 two-tailed significance level.

where (COMPUSTAT item numbers are in parentheses): R is raw daily percent-


age returns compounded over the fiscal year; COMP is net income
*+,&-./
adjusted for MKT-ADJ (!172# $!238); COMP is net income adjusted
01&-./
for FC-ADJ (!172# $!230); COMP is net income adjusted for
2345&-./
PENS-ADJ (!172#0.65 times $[!297!!298, if less than zero]);
COMP , COMP , and COMP are deflated by market value
*+,&-./ 01&-./ 2345&-./
of common equity at the previous fiscal year-end.
The results of estimating models (2a) to (2c) are reported in Table 2, panel B.
The adjusted R# using COMP is larger than the adjusted R# using net
*+,&-./
income and the difference in adjusted R# is significant at the 0.01 two-tailed
level, while the adjusted R# using COMP and COMP do not differ
01&-./ 2345&-./
from the adjusted R# using net income at conventional significance levels.!)
These results suggest that the marketable securities adjustment is the only
component of SFAS 130 ‘other comprehensive income’ that improves income’s
ability to summarize firm performance as reflected in stock returns.!!
Overall, these results suggest that the increase in the association between
income and returns observed when income is measured as COMP is driven
!$)

!) Bartov (1997) finds that the level of the cumulative foreign currency translation adjustment is
associated with returns, although this result is sensitive to the returns window used. His results are
difficult to compare with those reported here because (i) his tests use levels of, rather than changes in,
the cumulative foreign currency translation adjustment; (ii) his sample covers the years 1984 to 1990
while our sample covers the years 1994 and 1995; (iii) his returns windows differ from the returns
window used in this study; and (iv) he uses rank regression while we use OLS. As a sensitivity test, we
extend our sample back to 1984 and redefine our returns windows to be consistent with those in
Bartov, with qualitatively similar results.
!! Identical inferences are drawn if we augment a regression of returns on net income with each
component of SFAS 130 ‘other comprehensive income’ individually and examine the increase in
explanatory power for each augmented model.
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67 55

by the marketable securities adjustment. These results also suggest that, with the
exception of the marketable securities adjustment, the components of SFAS 130
‘other comprehensive income’ merely add noise to comprehensive income. One
possible explanation for these results is that the foreign currency translation and
minimum pension liability adjustments may involve more subjective estimates
than the marketable securities adjustment and thus may be more noisy than the
marketable securities adjustment.

3.3. Within-industry analysis

We now examine whether our full sample results differ between major indus-
try groups. There are two reasons for this analysis. First, income is used as
a summary measure of firm performance in many contracting and valuation
contexts. An examination of inter-industry differences can be used to draw
inferences regarding which measure of income, comprehensive income or net
income, is more appropriate to use as a summary measure of firm performance
for different industries.!# Second, an examination of inter-industry differences
will enable us to interpret and clarify our earlier results. For example, in tests
across industries, the marketable securities adjustment is the only component of
SFAS 130 ‘other comprehensive income’ that improves the association between
income and returns. Among industries, the marketable securities adjustment is
probably most closely related to a firm’s operations for financial sector firms
because these firms are in the business of managing financial assets. Thus, it is
possible that our results for the marketable securities adjustment are confined to
this sector.
To determine whether the results of our tests differ between major industry
groups, we repeat the tests summarized in Table 2 for each of the following
broad industry groups:
Financial: the 2,255 firm-years included in SIC 6000-6499 and SIC 6600-6999;
Manufacturing: the 4,826 firm-years included in SIC 2000-3999;
Merchandising: the 1,226 firm-years included in SIC 5000-5999;
Utilities: the 479 firm-years included in SIC 4900-4999;
Others: the 2,639 firm-years not included in the four preceding industry
classifications.
The results of within-industry estimates are summarized in Table 3. As
reported in panel A, in estimates of models 1a to 1c, the adjusted R# using

!# Also, as pointed out by Biddle et al. (1995), information on inter-industry differences is useful
because investors that face data acquisition and processing costs can reduce these costs by knowing
which measure of income is a better summary measure of firm performance for the industries they
follow.
56 D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

Table 3
Results of the estimation of models that test the association of net income and comprehensive
income with returns: Firms are classified by industry.!

Independent Row Manufac- Merchan-


Model" variable" label Financial turing dising Utilities Others

Panel A: Association of net income and comprehensive income with returns

(1a) NI Adj. R# 9.47 2.51 5.39 6.19 4.57


(1b) COMP Adj. R# 11.92 2.25 3.85 6.77 3.57
"#$!%
Vuong z-stat' (2.39)**% (!1.37) (!1.24) (0.18) (!1.82)
(1c) COMP Adj. R# 14.22 2.55 5.69 6.20 4.62
!$)
Vuong z-stat' (3.48)* (0.44) (1.43) (0.15) (0.49)

Panel B: Association of net income and component-based comprehensive income with returns

(2a) COMP Adj. R# 14.23 2.58 5.60 6.19 4.67


*+,&-./
Vuong z-stat' (3.48)* (1.76) (1.10) (0.13) (0.90)
(2b) COMP Adj. R# 9.52 2.46 5.48 6.20 4.60
01&-./
Vuong z-stat' (1.65) (0.58) (1.84) (0.55) (1.23)
(2c) COMP Adj. R# 9.39 2.52 5.39 6.18 4.49
2345&-./
Vuong z-stat' (!1.02) (1.39) (0.20) (!1.18) (1.52)

!The sample consists of all 1994 and 1995 firm-years that have COMPUSTAT and CRSP data
needed to calculate return, net income, and comprehensive income. Observations for which the
absolute value of the test variable related to net income or comprehensive income falls in the top
percentile of the test-variable distribution are eliminated from the sample. Financial are the 2,255
firm-years in SIC 6000-6499 and SIC 6600-6999; manufacturing are the 4,826 firm-years in SIC
2000-3999; merchandising are the 1,226 firm-years in SIC 5000-5999; utilities are the 479 firm-years
in SIC 4900-4999; others are the 2,639 firm-years in remaining SIC.
" Model (1a): R "! #" *NI ##
! ) ! ! !
Model (1b): R "! #" *COMP ##
! ) ! "#$!%"! !
Model (1c): R "! #" *COMP ##
! ) ! !$)"! !
Model (2a): R "! #" *COMP ##
! ) ! *+,&-./"! !
Model (2b): R "! #" *COMP ##
! ) ! 01&-./"! !
Model (2c): R "! #" *COMP ##
! ) ! 2345&-./"! !
Model variables are defined as follows: R is raw daily percentage returns compounded over the fiscal
year; NI is net income after extraordinary items and discontinued operations; COMP is change in
"#$!%
COMPUSTAT item !36, adjusted retained earnings, plus common stock dividends; COMP is as-if
!$)
SFAS 130 comprehensive income; OTH is ‘other comprehensive income’ equal to COMP minus
"#$!% "#$!%
NI; OTH is ‘other comprehensive income’ equal to COMP minus NI; MKT-ADJ is change in the
!$) !$)
balance of unrealized gains and losses on marketable securities; FC-ADJ is change in cumulative foreign
currency translation adjustment; PENS-ADJ is change in additional minimum pension liability in excess
of unrecognized prior service cost; COMP is net income adjusted for MKT-ADJ; COMP is
*+,&-./ 01&-./
net income adjusted for FC-ADJ; COMP is net income adjusted for PENS-ADJ. Missing
2345&-./
COMPUSTAT data for MKT-ADJ, FC-ADJ, and/or PENS-ADJ, are set equal to zero. All variables
except R are deflated by market value of common equity at the previous fiscal year-end.
'Adjusted R# are compared using the likelihood ratio test described in Vuong (1989). The row labeled
‘Vuong z-stat’ reports the z-statistic associated with the Vuong test.
% *[**] indicates that adjusted R# in the estimate of model (1a) differs from adjusted R# in the
estimate of model (1b),2,(2c) at the 0.01 [0.05] two-tailed significance level.
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67 57

COMP is larger than the adjusted R# using net income at the 0.01 two-tailed
!$)
significance level for financial sector firms, while in the other four industry
groups, the adjusted R# using COMP does not differ from the adjusted
!$)
R# using net income at conventional levels of significance. The tenor of the
results is unchanged using COMP instead of COMP . These findings
"#$!% !$)
suggest that comprehensive income is a better measure of firm performance than
other summary income measures only for financial sector firms. They also
suggest that the increase in the association between income and returns ob-
served in the full sample when income is measured as COMP is driven by
!$)
financial sector firms.
In the full sample, we find that the marketable securities adjustment is the
only component of SFAS 130 ‘other comprehensive income’ that improves
the association between income and returns. An examination of our within-
industry results indicates that this finding is driven by financial sector firms.
Specifically, as seen in panel B of Table 3, when we construct alternate measures
of income by adjusting net income individually for each component of SFAS 130
‘other comprehensive income’, only COMP for financial sector firms is
*+,&-./
more strongly associated with returns than net income at conventional signifi-
cance levels. This finding suggests that the increase in the association between
income and returns observed in the full sample when income is measured as
COMP is driven by the marketable securities adjustment for financial sector
!$)
firms.
To rule out the possibility that lower power stemming from smaller sample
sizes is leading to the insignificant results for the four non-financial sector
industry groups, we repeat the tests in this section for the pooled sample of 9,170
non-financial firm-years, with qualitatively similar results.

3.4. Can a lack of materiality explain the results?

We find that the marketable securities adjustment (MKT-ADJ) is the only


component of SFAS 130 ‘other comprehensive income’ (OTH ) that improves
!$)
the association between income and returns. One explanation for this result is
that the components of OTH other than MKT-ADJ are not material. The
!$)
descriptive data reported in Table 1 suggest this is a reasonable possibility.
Specifically, while the mean absolute value of MKT-ADJ is 0.010, the mean
absolute value of the foreign currency translation adjustment (FC-ADJ) and the
minimum pension liability adjustment (PENS-ADJ) are considerably smaller at
0.002 and 0.001, respectively. We therefore conduct a sensitivity analysis where-
in we estimate models (2a) to (2c) when the sample is limited to the decile with
the largest absolute magnitude of the relevant component adjustment. An
examination of descriptive data for this restricted sample, reported in the ‘all
industries’ column in panel A of Table 4, indicates that the relative lack of
58
Table 4
Results of tests of the association of net income and comprehensive income with returns: Firm-years with large comprehensive income adjustments!
Panel A

All industries Financial sector Non-financial sectors

Sample Mean absolute Sample Mean absolute Sample Mean absolute


Variable" size value size value size value

OTH 1142 0.185 627 0.175 515 0.194


"#$!%
OTH 1142 0.101 853 0.107 289 0.084
!$)
MKT-ADJ 1142 0.093 1002 0.095 140 0.080
FC-ADJ 1142 0.018 51 0.021 1091 0.018
PENS-ADJ 936 0.009 87 0.005 849 0.009

Models Independent Adj. R# of model


compared" variable" Adj. R# of model (1a)' compared to (1a)' Vuong z-stat'

Panel B: Results of estimation of models: All industries


(1a) to (1b) COMP 2.97 1.73 !1.21
"#$!%
(1a) to (1c) COMP 8.62 13.02 1.85
!$)
(1a) to (2a) COMP 8.96 19.49 5.37*%
*+,&-./
(1a) to (2b) COMP 4.76 4.70 !0.17
01&-./
(1a) to (2c) COMP 3.96 3.64 !1.06
2345&-./
Panel C: Results of estimation of models: Financial sector
(1a) to (1b) COMP 8.18 18.05 3.32*
"#$!%
(1a) to (1c) COMP 7.62 23.25 5.93*
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

!$)
(1a) to (2a) COMP 8.89 23.61 5.76*
*+,&-./
(1a) to (2b) COMP 12.51 17.81 1.76
01&-./
(1a) to (2c) COMP 13.65 11.10 !0.96
2345&-./
Panel D: Results of estimation of models: Non-financial sectors

(1a) to (1b) COMP 2.93 0.00 !1.23


"#$!%
(1a) to (1c) COMP 10.39 10.64 0.11
!$)
(1a) to (2a) COMP 11.66 15.55 1.45
*+,&-./
(1a) to (2b) COMP 4.86 4.75 !0.26
01&-./
(1a) to (2c) COMP 3.42 3.21 !0.75
2345&-./
!The original sample of 11,425 firm-years is described in the notes to Table 1. The ‘all industries’ samples in this table consist of firm-years in the original
sample that fall in the decile with the largest absolute magnitude of OTH (OTH ; MKT-ADJ; FC-ADJ; PENS-ADJ); there are only 936
"#$!% !$)
observations in the PENS-ADJ sample because there are only 936 non-zero values of PENS-ADJ in the original sample. The ‘financial sector’
(‘non-financial sectors’) samples in this table consist of the firm-years in the ‘all industries’ samples that are in (that are not in) SIC 6000-6499 and SIC
6600-6999.
" Model (1a): R "! #" *NI ##
! ) ! ! !
Model 1b: R "! #" *COMP ##
! ) ! "#$!%"! !
Model 1c: R "! #" *COMP ##
! ) ! !$)"! !
Model 2a: R "! #" *COMP ##
! ) ! *+,&-./"! !
Model 2b: R "! #" *COMP ##
! ) ! 01&-./"! !
Model 2c: R "! #" *COMP ##
! ) ! 2345&-./"! !
Model variables are defined as follows: R is raw daily percentage returns compounded over the fiscal year; NI is net income after extraordinary items and
discontinued operations; COMP is change in COMPUSTAT item !36, adjusted retained earnings, plus common stock dividends; COMP is as-if
"#$!% !$)
SFAS 130 comprehensive income; OTH is ‘other comprehensive income’ equal to COMP minus NI; OTH is ‘other comprehensive income’
"#$!% "#$!% !$)
equal to COMP minus NI; MKT-ADJ is change in the balance of unrealized gains and losses on marketable securities; FC-ADJ is change in
!$)
cumulative foreign currency translation adjustment; PENS-ADJ is change in additional minimum pension liability in excess of unrecognized prior service
cost; COMP is net income adjusted for MKT-ADJ; COMP is net income adjusted for FC-ADJ; COMP is net income adjusted for
*+,&-./ 01&-./ 2345&-./
PENS-ADJ. All variables except R are deflated by market value of common equity at the previous fiscal year-end.
'The column labeled ‘Adj. R# of model (1a)’ reports adjusted R# for the estimate of model (1a). The column labeled ‘Adj. R# of model compared to (1a)’
reports adjusted R# for the estimate of the model that model 1a is being compared to. These adjusted R# are compared using the likelihood ratio test
described in Vuong (1989). The column labeled ‘Vuong z-stat’ is the z-statistic associated with the Vuong test.
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

%*indicates that adjusted R# for the estimate of model (1a) differs from adjusted R# for the estimate of the model that model (1a) is being compared to at the
0.01 two-tailed significance level.
59
60 D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

materiality for FC-ADJ and PENS-ADJ is alleviated, with the mean absolute
value of FC-ADJ (PENS-ADJ) increasing from 0.002 (0.001) to 0.018 (0.009).Re-
sults for restricted-sample estimates of models 2a to 2c, reported in rows three to
five of panel B of Table 4, are qualitatively similar to those reported in Table 2.
That is, MKT-ADJ is the only component adjustment that improves the
association between income and returns.
Further, our within-industry results for COMP may be attributable to
!$)
lack of materiality in the non-financial sectors. For example, while the mean
absolute value of OTH in the financial sector is 0.043, it is only 0.005 in the
!$)
non-financial sectors. We therefore conduct a sensitivity analysis wherein we
estimate models 1a and 1c separately for financial sector firms and non-financial
sector firms that fall in the upper decile of the absolute value of OTH
!$)
distribution. As reported in the ‘non-financial sectors’ column in panel A of
Table 4, this procedure alleviates the relative lack of materiality for OTH in
!$)
the non-financial sectors, with the mean absolute value increasing from 0.005 to
0.084. Results for restricted-sample estimates of models (1a) and (1c), reported in
row two of panels C and D of Table 4, are qualitatively similar to those reported
in Table 3. That is, the adjusted R# using COMP is larger than the adjusted
!$)
R# using net income at conventional significance levels only in the financial
sector. The tenor of these results is unchanged using COMP instead of
"#$!%
COMP .
!$)
We also estimate models (2a) to (2c) separately for financial sector firms and
non-financial sector firms after the sample is limited to the decile with the largest
absolute magnitude of the relevant component adjustment. Our results are
qualitatively similar to the results reported in Table 3. That is, as reported in
rows 3 to 5 of panels C and D of Table 4, the only component adjustment that
improves the association between income and returns is MKT-ADJ for financial
sector firms.
The qualitative similarity of the results reported in Table 4 and those reported
earlier suggests that a lack of materiality probably does not explain earlier
results.

3.5. Implications of the results

Overall, our results do not support the claim that income measured on
a comprehensive basis is a better measure of firm performance than other
summary income measures. Our results also suggest that the foreign currency
translation and minimum pension liability adjustments merely add noise to
comprehensive income. Further, our results for non-financial industry groups
raise questions about whether it is appropriate to require uniform comprehens-
ive income disclosures for all industries.
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67 61

The fact that significant results for the marketable securities adjustment are
confined to financial sector firms appears to be reasonable because, among
industries, only financial sector firms are primarily in the business of managing
financial assets. This ex post observation suggests that to improve the ability of
comprehensive income to summarize firm performance, the Financial Ac-
counting Standards Board should focus on items that are closely related to
a firm’s operations.

4. Tests of association with price and with future operating cash flows and income

4.1. Tests of the association of alternate income measures with the market value
of equity

Due to both econometric and theoretical problems with the returns model,
Kothari and Zimmerman (1995) suggest that researchers should use additional
models in their empirical analysis, such as the price model, to draw more
definitive inferences.!$ Accordingly, we estimate price level models in which the
dependent variable is the market value of equity and the independent variable is
alternately net income, COMP , or COMP .!% Although the models tested
"#$!% !$)
here do measure the association between income and firm performance as
reflected in stock price, they “do not measure information arrival over a period.
The dependent variable, price, is not a measure of the impact of information
arriving in the current period. In an efficient market, the impact of information
over a period is measured by stock returns2” (Kothari and Zimmerman, 1995)
(p. 184).
The results of the preceding analysis, reported in the column headed ‘Price ’ in
!
Table 5, indicate that both COMP and COMP are less strongly asso-
"#$!% !$)
ciated with the market value of equity than net income. In particular, the
adjusted R# using either COMP or COMP is smaller than the adjusted
"#$!% !$)

!$ Kothari and Zimmerman (1995) (p. 183) suggest that “[t]he findings in this paper do not
suggest using either price or return models exclusively, because both have serious econometric
problems and both have important deviations from the underlying theoretical model. Future studies
can be enriched by testing for sensitivity to the functional form and by incorporating the relative
strengths and weaknesses of alternative specifications. Using the price model, perhaps in addition to
the return model, could permit more definitive inferences.”
!% Details of sample selection, models, and variable definitions are provided in Table 5. In these
tests, sample size is 11,348 firm-years. The smaller number of firm-years in these tests is primarily the
result of missing COMPUSTAT data.
62 D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

Table 5
Results of the estimation of models that test the association of net income and comprehensive
income with market value of equity and with future cash flows/income!

Dependent variable"

Independent variable(s)" Row label Price Cash flow Net income


! !"! !"!
NI Adj. R# 36.42 16.54 30.69
)
COMP Adj. R# 28.11 11.43 22.35
"#$!%")
Vuong z-stat' (!6.68)*% (!4.12)* (!5.59)*
COMP Adj. R# 33.56 13.04 27.06
!$)")
Vuong z-stat' (!8.64)* (!3.24)* (!4.23)*
NI , BV Adj. R# 52.77
) )
COMP , BV Adj. R# 50.87
"#$!%") )
Vuong z-stat' (!4.49)*
COMP , BV Adj. R# 52.08
!$)") )
Vuong z-stat' (!5.62)*

!For the models reported in the ‘Price ’ column, the sample consists of all 1994 and 1995 firm-years
!
that have COMPUSTAT data needed to calculate market value of equity, net income, comprehens-
ive income, and book value of equity. Observations for which the absolute value of an independent
falls in the top percentile of the variable distribution are eliminated from the sample. Sample size is
11,348 firm-years. For the models reported in the ‘Cash Flow ’ and ‘Net Income ’ columns, the
!"! !"!
sample consists of all 1994, 1995 and 1996 firm-years that have COMPUSTAT data needed to
calculate operating cash flows, net income, and comprehensive income. Observations for which the
absolute value of an independent variable falls in the top percentile of the variable distribution are
eliminated from the sample. Sample size is 8,893 firm-years.
"Variable definitions are as follows: NI is net income after extraordinary items and discontinued
operations; COMP is change in COMPUSTAT item !36, adjusted retained earnings, plus
"#$!%
common stock dividends; COMP is as-if SFAS 130 comprehensive income; BV is book value of
!$)
equity at fiscal year-end; PRICE is market value of common equity at fiscal year-end; and CASH
FLOW is net cash flow from operating activities (COMPUSTAT item !308). Variables are deflated
by number of shares of common stock outstanding at fiscal year-end adjusted for stock splits and
stock dividends. The models estimated in the ‘Price ’ column are: PRICE "! #" *NI #! ;
! ! ) ! ! !
PRICE "! #" *COMP #! ; PRICE "! #" *COMP #! ; PRICE "! #" *NI
! ) ! "#$!%"! ! ! ) ! !$)"! ! ! ) ! !
#" *BV #! ; PRICE "! #" *COMP #" *BV #! ; PRICE "! #" *COMP
# ! ! ! ) ! "#$!%"! # ! ! ! ) ! !$)"!
#" *BV #! . The models estimated in the ‘Cash Flow ’ column are: CASH FLOW "
# ! ! !"! !"!
! #" *NI #! ; CASH FLOW "! #" *COMP #! ; CASH FLOW "! #
) ! ! ! !"! ) ! "#$!%") ) !"! )
" *COMP #! . The models estimated in the ‘Net Income ’ column are the same as those
! !$)"! ! !"!
estimated in the ‘Cash Flow ’ column except that the dependent variable is NI .
!"! !"!
'Model adjusted R# are compared using the likelihood ratio test described in Vuong (1989). The row
labeled ‘Vuong z-stat’ reports the z-statistic associated with the Vuong test.
%* Indicates that adjusted R# in the estimate of the model using NI as an independent variable differs
from adjusted R# in the estimate of the model using COMP (COMP ) as an independent
"#$!% !$)
variable at the 0.01 two-tailed significance level.
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67 63

R# using net income at the 0.01 two-tailed significance level. These results do not
support the claim that income measured on a comprehensive basis is a better
measure of firm performance than other summary income measures.
Based on recent evidence (e.g., Berger et al., 1996; Burgstahler and Dichev,
1997) that price/earnings models can be misspecified due to the omission of
book value of equity, we repeat the preceding analysis after augmenting the
models with book value of equity. As reported in the column headed ‘Price ’ in
!
Table 5, we again find that the adjusted R# using either COMP or
"#$!%
COMP is significantly smaller than the adjusted R# using net income.
!$)
However, the magnitudes of the differences in adjusted R# are smaller using the
augmented model.

4.2. Tests of the association of alternate income measures with future


operating cash flows and income

Firm performance should be reflected in future operating cash flows and


income as well as in stock returns (Dechow et al., 1998). Thus, if comprehensive
income is a better measure of firm performance than other summary income
measures, then future operating cash flows and income should be more strongly
associated with comprehensive income than with net income. We estimate
cross-sectional regressions to test this prediction because our sample is con-
strained to just two years. These cross-sectional models are specified such that
the dependent variable is alternately operating cash flows or net income in year
t#1 for a given firm and the independent variable is alternately net income,
COMP , or COMP in year t for the corresponding firm (t"1994,
"#$!% !$)
1995).!&
The results of estimating these models are reported in the two right-hand-
most columns in Table 5. For each dependent variable specification, the ad-
justed R# using net income significantly exceeds the adjusted R# using either
COMP or COMP . The fact that future operating cash flows and net
"#$!% !$)
income are more strongly associated with net income than they are with
comprehensive income again weakens the case for comprehensive income as
a better measure of firm performance than net income.
In their tests of whether current profitability helps forecast future profitability
and earnings, Fama and French (1997) include dividends, market to book ratio,

!& Details of sample selection, models, and variable definitions are provided in Table 5. In these
tests, sample size is 8,893 firm-years. The smaller number of firm-years in these tests is primarily the
result of missing COMPUSTAT data.
64 D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

and leverage as control variables. As a sensitivity test, we repeat our analysis


including the same set of control variables and obtain qualitatively similar
results to those reported in Table 5.

5. Summary and implications

This study’s major findings are as follows. We find no clear evidence that
comprehensive income is on average more strongly associated with returns than
net income, and comprehensive income is less strongly associated with the
market value of equity and predicts future operating cash flows and income
worse than net income. Further, the marketable securities adjustment is the only
component of SFAS 130 ‘other comprehensive income’ that improves the
association between income and returns. Additional analysis indicates that the
latter finding is driven by financial sector firms.
Our results have the following implications. First, our results do not sup-
port the claim that income measured on a comprehensive basis is a better
measure of firm performance than other summary income measures. Second,
with the exception of the marketable securities adjustment, the components of
SFAS 130 ‘other comprehensive income’ merely add noise to comprehensive
income. Thus, the Financial Accounting Standards Board may want to re-
examine the requirement to include these items in comprehensive income.
Finally, our insignificant results for non-financial industries bring into question
the usefulness of mandating uniform comprehensive income disclosures for all
industries.
The inferences drawn in this study are subject to an important caveat.
Although we do provide evidence on which comprehensive income adjust-
ments improve income’s ability to summarize firm performance, this
evidence cannot form the basis for deciding whether these items should neces-
sarily be included in comprehensive income, or even disclosed in the financial
statements, for at least two reasons. First, to decide whether an item should
be disclosed, a cost-benefit analysis, considering issues such as whether
the component is readily available from alternate, non-financial statement
sources, would have to be carried out. Second, as discussed previously, even
if it is decided that an item should be disclosed, it is not clear whether it
matters if the item is included in comprehensive income or disclosed elsewhere
in the financial statements (e.g., as a direct adjustment to equity). Evi-
dence on this latter issue will have to wait until SFAS 130 disclosures are
available for fiscal years beginning after 15 December 1997 and will entail
a substantially different research design than the association study used in this
paper.
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67 65

Appendix A. Example of SFAS 130 financial statement disclosures

References
Accounting Principles Board, 1966. APB Opinion no. 9: Reporting the results of operations.
American Institute of Certified Public Accountants, New York.
Ahmed, A.S., Takeda, C., 1995. Stock market valuation of gains and losses on commercial banks’
investment securities: An empirical analysis. Journal of Accounting and Economics 20, 207—225.
Ali, A., Zarowin, P., 1991. Permanent versus transitory components of annual earnings and
estimation error in earnings response coefficients. Journal of Accounting and Economics 15,
249—264.
Ali, A., Zarowin, P., 1992. The role of earnings levels in annual earnings—returns studies. Journal of
Accounting Research 30, 286—296.
American Accounting Association Financial Accounting Standards Committee, 1997. An issues
paper on comprehensive income. Accounting Horizons 11, 120—126.
Association for Investment Management and Research, 1993. Financial reporting in the 1990s and
beyond. Association for Investment Management and Research, Charlottesville, VA.
Barth, M.E., 1994. Fair value accounting: Evidence from investment securities and the market
valuation of banks. The Accounting Review 69, 1—25.
Barth, M.E., Landsman, W.R., Wahlen, J.M., 1995. Fair value accounting: Effect on banks’ earnings
volatility, regulatory capital, and value of contractual cash flows. Journal of Banking and Finance
19, 577—605.
Barth, M.E., Beaver, W.H., Landsman, W.R., 1996. Value-relevance of banks’ fair value disclosures
under SFAS 107. The Accounting Review 71, 513—537.
Bartov, E., 1997. Foreign currency exposure of multinational firms: Accounting measures and
market valuation. Contemporary Accounting Research 14, 623—652.
Beresford, D.R., Johnson, L.T., Reither, C.L., 1996. Is a second income statement needed?, Journal of
Accountancy (April), 69—72.
Berger, P.G., Ofek, E., Swary, I., 1996. Investor valuation of the abandonment option. Journal of
Financial Economics 42, 257—287.
66 D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67

Bernard, V., Schipper, K., 1994. Recognition and disclosure in financial reporting. Unpublished
paper, University of Chicago, Chicago, IL.
Biddle, G.C., Seow, G.S., Siegel, A.F., 1995. Relative versus incremental information content.
Contemporary Accounting Research 12, 1—23.
Black, F., 1993. Choosing accounting rules. Accounting Horizons 7, 1—17.
Bowman, R.G., 1980. The debt equivalence of leases: An empirical investigation. The Accounting
Review 55, 237—253.
Brief, R.P., Peasnell, K.V. (Eds.), 1996. Clean surplus: A link between accounting and finance.
Garland Publishing, New York.
Burgstahler, D.C., Dichev, I.D., 1997. Earnings, adaptation and equity value. The Accounting
Review 72, 187—215.
Collins, D.W., Kothari, S.P., 1989. An analysis of intertemporal and cross-sectional determinants of
earnings response coefficients. Journal of Accounting and Economics 11, 143—181.
Dechow, P.M., 1994. Accounting earnings and cash flows as measures of firm performance: The role
of accounting accruals. Journal of Accounting and Economics 18, 3—42.
Dechow, P.M., Kothari, S.P., Watts, R.L., 1998. The relation between earnings and cash flows.
Journal of Accounting and Economics, forthcoming.
Dhaliwal, D.S., 1986. Measurement of financial leverage in the presence of unfunded pension
obligations. The Accounting Review 61, 651—661.
Easton, P.D., Harris, T.S., 1991. Earnings as explanatory variables for returns. Journal of Ac-
counting Research 29, 19—36.
Eccher, E.A., Ramesh, K., Thiagarajan, S.R., 1996. Fair value disclosures by bank holding com-
panies. Journal of Accounting and Economics 22, 79—117.
Fama, E.F., French, K.R., 1997. Forecasting profitability and earnings. Unpublished paper, Univer-
sity of Chicago, Chicago, IL.
Financial Accounting Standards Board, 1984. Concepts statement no. 5: Recognition and measure-
ment in financial statements of business enterprises. Financial Accounting Standards Board,
Stamford, CT.
Financial Accounting Standards Board, 1985. Concepts statement no. 6: Elements of financial
statements. Financial Accounting Standards Board, Stamford, CT.
Financial Accounting Standards Board, 1997. Statement of financial accounting standards no. 130:
Reporting comprehensive income. Financial Accounting Standards Board, Norwalk, CT.
Financial Accounting Standards Board, 1998. Statement of financial accounting standards no. 133:
Accounting for derivative instruments and hedging activities. Financial Accounting Standards
Board, Norwalk, CT.
Foster, N., Hall, N.L., 1996. Reporting comprehensive income. The CPA Journal (October),
16—19.
International Accounting Standards Committee, 1996. Exposure draft: Presentation of financial
statements. International Accounting Standards Committee, London.
Johnson, L.T., Reither, C.L., Swieringa, R.J., 1995. Toward reporting comprehensive income.
Accounting Horizons 9, 128—137.
Johnson, L.T., Swieringa, R.J., 1996. Derivatives, hedging, and comprehensive income. Accounting
Horizons 10, 109—122.
Kothari, S.P., 1992. Price earnings regressions in the presence of prices leading earnings: Implica-
tions for earnings response coefficients. Journal of Accounting and Economics 15, 173—202.
Kothari, S.P., Zimmerman, J.L., 1995. Price and return models. Journal of Accounting and
Economics 20, 155—192.
Nelson, K., 1996. Fair value accounting for commercial banks: An empirical analysis of SFAS no.
107. The Accounting Review 71, 161—182.
O’Hanlon, J.F., Pope, P.F., 1997. The value-relevance of dirty surplus accounting flows. Unpub-
lished paper, Lancaster University, Lancaster.
D. Dhaliwal et al. / Journal of Accounting and Economics 26 (1999) 43–67 67

Ohlson, J.A., Shroff, P., 1992. Changes versus levels in earnings as explanatory variables for returns:
Some theoretical considerations. Journal of Accounting Research 30, 210—226.
Subramanyam, K.R., 1996. The pricing of discretionary accruals. Journal of Accounting and
Economics 22, 249—281.
Vuong, Q.H., 1989. Likelihood ratio tests for model selection and non-nested hypotheses. Econo-
metrica, 307—333.

You might also like