You are on page 1of 6

A-PDF N-up Page Demo. Purchase from www.A-PDF.

com to remove the watermark

The Predictive Power of Comprehensive The Predictive Power of Comprehensive


Income and Its Individual Components under Income and Its Individual Components under
IFRS IFRS
Paul Pronobis, Center for Financial Reporting & Consolidation, HHL – Leipzig School of Management,
Germany, E-mail: paul.pronobis@hhl.de Abstract
Henning Zülch, Chair of Accounting and Auditing, HHL – Leipzig School of Management, Germany, E- This study examines the predictive power of comprehensive income and its individual components within
mail: henning.zuelch@hhl.de the homogenous institutional setting of German IFRS firms. The results could be relevant for the standard
setters IASB and FASB and their joint project “Financial Statement Presentation”. We find no evidence that
comprehensive income has a superior predictive power for future firm operating performance than net
income. Further, we fail to find significant incremental predictive power of aggregated or individual com-
ponents of other comprehensive income for subsequent period’s firm operating performance. The actuarial
Further contact information: HHL – Leipzig Graduate School of Management, Jahnallee 59, D-04109 gains and losses on defined benefit pension obligations even seem to merely add noise for the prediction of
subsequent period’s net income and of subsequent period’s comprehensive income. In contrast, our analys-
Leipzig, Germany; phone: +49 (0) 341 9851 701; fax: +49 (0) 341 9851 702.
es indicate that other comprehensive income components seem to have incremental predictive power
beyond one period. Finally, we find that the predictive power of net income and comprehensive for future
firm operating performance has deteriorated as a consequence of the IASB’s recent initiatives and actions.

Keywords: accounting, income, earnings, Germany, IAS, IFRS, predictability, predictive power, earnings
quality, net income, operating cash flows, other comprehensive income, OCI items, comprehensive income,
financial statement presentation

1. Introduction Other comprehensive income items are currently


Under current IFRS regulation, companies have to presented in the statement of changes in equity.
book certain transactions directly into equity instead However, the IASB is conducting a joint project with
of into the income statement. These transactions are the FASB called “Financial Statement Presentation”
often referred to as ‘dirty surplus (non-all-inclusive) which already resulted in a revised IAS 1 ‘Presenta-
components of income for the period’. They mainly tion of Financial Statements’ and will alter the loca-
represent fair value changes in balance sheet ac- tion of other comprehensive income presentation as
counts, and therefore, are transitory in nature. The follows: For annual periods from 2009 onwards,
dirty surplus components of income for the period other comprehensive income should be presented as
are better known as other comprehensive income part of a new ‘bottom line’ within the income state-
(OCI) and include such items as foreign currency ment. Thus, the other comprehensive income items
translation adjustments (IAS 21.39[c]), unrealized are added and subtracted, respectively, from net
gains (losses) on re-measuring available-for-sale income. The resulting balance is labeled compre-
(AfS) financial assets (IAS 39.55[b]), effective por- hensive income and represents an ‘all-inclusive
tion of gains and losses on hedging instruments in a number of income for the period’. Alternatively, the
cash flow hedge (IAS 39.95[a]), effective portion of reconciliation from net income to comprehensive
gains or losses on hedging instruments of a net in- income can be presented in the notes.
vestment in a foreign operation (IAS 39.102), reval- From the point of view of economic research, the
uation of property, plant and equipment (IAS 16.39), format of presentation of accounting information is
revaluation of intangible assets (IAS 38.85), actuari- irrelevant as long as the same items are included.
al gains (losses) on defined benefit pension obliga- However, empirical and experimental accounting
tions (IAS 19.93B), and deferred tax effects on all research does show that presentation format might
items reported in OCI (IAS 12.61). influence investors’ decisions (e. g. Hirst and
Hopkins [1998], Maines and McDaniel [2000]).
Building upon this evidence, the usefulness of pre-

1 2

Electronic copy available at: http://ssrn.com/abstract=1576384 Electronic copy available at: http://ssrn.com/abstract=1576384

senting comprehensive income instead of net in- as Worldscope and Hoppenstedt lack detail and position which can directly be booked into equity are 2.2 The financial statement presentation
come as the bottom line of the income statement has quality with regard to (other) comprehensive in- gains and losses on the re-measuring of available- project of IASB and FASB
been the research focus of many empirical studies, come reporting. Since for some of the measures for-sale (AfS) securities. This category comprises all Aiming at developing a ‘conceptually robust‘ ac-
especially in the US American territory (e. g. Dha- examined in this research it is necessary to use in- non-derivative financial instruments that are not counting standard which addresses disclosure issues
liwal, Subramanyan, and Trezevant [1999]). In the formation of the previous or the following period, classified as either held-to-maturity investments, in the statement of comprehensive income the IASC,
US a similar change with regard to the presentation data of the years 1997 and 2008 is also included. loans and receivables or those securities noted as predecessor of the IASB, initiated the project “Re-
of (other) comprehensive was caused by the is- Our results indicate no superior predictive power of fair value through profit or loss (FVTPL). porting Financial Performance” which was re-
suance of SFAS no. 130, Reporting Comprehensive comprehensive income over net income for future The following two OCI items refer to derivative fi- named “Reporting Comprehensive Income” in 2003
Income, in 1997. These empirical research studies firm operating performance. In addition, we fail to nancial instruments. Out of the three categories of and “Financial Statement Presentation” in 2006,
examine the claim that income measured on a com- find significant incremental predictive power of derivatives, which are fair value hedges, cash flow respectively. This project is now being conducted in
prehensive basis is a better measure of firm perfor- aggregated or individual components of OCI for hedges and hedges in foreign currency exposures of cooperation with the FASB in order to align the two
mance than other summary income measures (e. g. subsequent period’s firm operating performance. net investment in foreign operations, the effective accounting systems.
net income). The above claim is primarily investi- The actuarial gains and losses on defined benefit portions of gains and losses of the last two categories Two driving forces for this project have been the
gated by testing whether comprehensive income or pension obligations even seem to merely add noise are booked directly into equity. increasing significance of OCI items as well as the
net income better summarizes firm performance as for the prediction of subsequent period’s net income Following IAS 16.31-42 and IAS 38.75-87, compa- growing use of pro-forma figures in performance
reflected in stock prices (so called association study). and of subsequent period’s comprehensive income, nies are given the option to apply a revaluation reporting of companies (Thinggaard, Wagenhofer,
The evidence is mixed. respectively. This might be driven by the fact that model for subsequent measurement of property, Evans, Gebhardt, Hoogendorn, Marton, di Pietro,
The current study wants to add empirical evidence under current regulation future plan amendments plant and equipment and intangible assets, respec- Mora, and Peasnell [2006]).
to the usefulness of (other) comprehensive income and future years of service are not anticipated re- tively. Within these revaluation models, those value The IASB’s objectives with regard to the Financial
disclosures by investigating a less researched field of gardless of the probability of occurrence. In contrast, increases above book value that do not equalize Statement Presentation project can be summed up
study within the homogeneous institutional back- OCI components seem to have incremental predic- previous periods’ impairments booked income in one main objective: To further increase the deci-
ground of German IFRS firms: the predictive power tive power beyond one period. In conclusion, the statement-related have to be booked directly into sion usefulness of financial statements by improving
of comprehensive income and its individual compo- incremental predictive power of OCI components equity. the predictive power of income numbers. The pre-
nents, namely, net income and other comprehensive depends on the specific prediction period under Next, actuarial gains and losses on defined benefit dictive power / predictability of income numbers is
income. The predictive power of income numbers is examination. pension obligations are components of other com- often referred to as an attribute of the so called
an attribute of high relevance for analysts as it re- Further, we find that the predictive power of net prehensive income. Since 2004 the IASB allows “earnings quality” (Dechow and Schrand [2004]).
duces their forecast risks. In addition, the IASB and income and comprehensive for future firm operating companies to book these actuarial gains and losses In order to realize this objective, the IASB divided its
FASB are prominently referring to the importance of performance seems to have deteriorated as a conse- directly into equity for the period the specific gains project into three phases. Phase A was dealing with
financial statements to assist their users in predict- quence of the IASB’s recent initiatives and actions. and losses are incurred. general questions regarding the income statement
ing the entity’s future cash flows and, in particular, Structural breaks of the over time patterns indicate Finally, the deferred tax effects corresponding to the presentation under IFRS. It was completed with the
their timing and certainty. Finally, the examination that the worsening of predictive power seems to be OCI items presented before have to be displayed in issuance of IAS 1 (rev. 2007). This revised standard
of predictive power provides direct evidence on the driven by main changes of the standards. one aggregate position within equity. In the further has to be applied by companies for business years
relation between accounting information and future course of this study the deferred tax effects will not beginning on or after January 2009 (earlier adop-
firm operating performance. 2. Background be considered, as it is not possible to allocate them tion permitted). While regulations for valuation of
Thus, at first, we evaluate the predictive power of the to the underlying transaction. income components remained unchanged, the pres-
different income numbers with regard to future 2.1 Current reporting requirements entation requirements were altered significantly.
operating cash flows. Second, we evaluate the pre- Under current IFRS regulation, companies have to Table 1: The revised IAS 1 standard requires IFRS adopters
dictability of comprehensive income items for future book certain transactions directly into equity. An Overview of OCI items under IFRS either to report other comprehensive income togeth-
net income and future comprehensive income. Over overview of items leading to such a presentation Position Standard er with components of profit or loss in a joint com-
and above, we examine the predictive power of the within the financial statements is given in Table 1 Foreign currency translation adjustments IAS 21.39(c) prehensive income statement (= single statement
different income numbers beyond one period. Final- below. Unrealized gains (losses) on re-measuring available- approach) or to remain with a two statement ap-
IAS 39.55(b)
ly, the results are segmented by time period in order The first position that is of relevance for most com- for-sale (AfS) financial assets proach (IAS 1.81 [rev. 2007]). However, in case us-
to analyze the stability of predictive power over time. panies with foreign operations is dealing with gains Effective portion of gains and losses on hedging ers are not following the one statement solution,
IAS 39.95(a)
A sound analysis of various criteria of the different instruments in a cash flow hedge they have to report a statement of comprehensive
and losses arising from translating the financial
income numbers is necessary due to the lack of a statements of these operations. Translation differ-
Effective portion of gains or losses on hedging
IAS 39.102 income (beginning with profit or loss of the period
instruments of a net investment in a foreign
conclusive theoretical standard of evaluation (De- ences arise if foreign subsidiaries use another func- and comprising all OCI items) in equal prominence
Revaluation of property, plant and equipment IAS 16.39
chow and Schrand [2004], p. 10 cont.). tional currency than the functional currency of the to the separate income statement. This increased
Revaluation of intangible assets IAS 38.85
We hand-collected (other) comprehensive income group. In case of similar functional currencies, prominence of comprehensive income marks a step
data for all companies listed in the HDAX in the Actuarial gains (losses) on defined benefit pension towards clean surplus accounting.
translation differences are booked into the income obligations
IAS 19.93B
period of 1998 to 2007. A hand-collection was re- statement rather than directly into equity. The next In Phase B of the project, the standard setters cur-
quired due to the fact that common databases such rently propose a single statement of comprehensive

3 4
income as the sole presentation format (DP Finan- comprehensive income and its individual compo- As can be seen above, the study conducted by Dha- 3. Hypotheses development
cial Statement Presentation 3.24-41). This step nents, especially with regard to IFRS. This is surpris- liwal, Subramanyan, and Trezevant (1999) con-
should enhance the decision usefulness of perfor- ing as the predictive power of income numbers also cludes that net income predicts future operating 3.1 Conceptual network
mance reporting. Again, this change in presentation represents a highly important attribute of earnings cash flows and income better than comprehensive Following an approach introduced by Libby (1981) a
should lead to an improvement of the predictive quality (e.g. Francis, LaFond, Olsson, and Schip- income. Biddle and Choi (2006) find that no income conceptual network is used to transfer the assumed
power of income numbers. Phase C shall align the per [2004]). First, predictive power is prominently definition dominates clearly with regard to its re- theoretical relationship into operational definitions.
interim financial statements to the proposed mentioned within the frameworks of IFRS and US spective predictive power, while Kanagaretnam, The results are pictured in Figure 1.
changes in performance reporting. GAAP. For example, according to F.15 “the econom- Mathieu, and Shehata (2009) observe a better pre- According to the arguments of the standard setters
ic decisions that are taken by users of financial dictive power of current period’s comprehensive (see above) and as represented by link 1 in Figure 1,
3. Literature review statements require an evaluation of the ability of an income over net income for future net income, fu- the increased prominence of (other) comprehensive
Most academic studies dealing with (other) compre- enterprise to generate cash and cash equivalents and ture comprehensive income and future operating income shall improve the predictability of future
hensive income are investigating the relevance of of the timing and certainty of their generation”. cash flows. Choi and Zang (2006) as well as Choi, firm operating performance. Due to the fact that the
comprehensive income compared to net income for Second, analysts need to predict future (operating) Das, and Zang (2007) confirm this superiority of concept of future firm operating performance can-
the prediction of future stock prices. This aspect cash flows or income for their valuation models. A comprehensive income as a predictor for future net not be measured directly, an operational definition
under examination is often referred to as “value high accuracy in prediction and low risk of revision, income. Although they do not find statistically sig- is derived to be the ability to predict future period’s
relevance” and also represents an attribute of earn- respectively, greatly reduces the forecast risks of nificant implications of one specific OCI component operating cash flows and earnings (e. g. net income
ings quality (Dechow and Schrand [2004]). A com- their models. Third, examining the predictive power for the prediction of future net income, Choi and and comprehensive income) as prominent variables
prehensive overview of these value relevance studies of income numbers directly informs about the rela- Zang (2006) find out that the prefix of the OCI val- of firm performance (expressed by link 2).
is given by Thinggaard, Wagenhofer, Evans, Geb- tion between accounting information and future ues is exerting predictive power. They detect a sig- The theoretical claim about an increased promi-
hardt, Hoogendorn, Marton, di Pietro, Mora, and firm operating performance. The results may serve nificant association between large positive OCI val- nence of (other) comprehensive income in reporting
Peasnell (2006) who are listing studies performed as a robustness test for the value relevance studies. ues and an increase in next period’s net income, and is transformed (via link 3) into an operational defini-
worldwide until 2006. More recently conducted Table 2 presents previous studies on the predictive vice versa. They explain this association with the tion which assumes that financial statement users
studies are mentioned by Goncharov and Hodg- power of comprehensive income and its individual accessed discretion of managers in choosing the replace net income by (other) comprehensive in-
son (2008). Although the results in this field of re- components. As can be seen, seven studies dealt timing of recognizing OCI components depending come as the key earnings measure in their analyses.
search are varied, the overall indication is that net with this issue within the recent past. Two of these upon the companies’ underlying performances. The Now an empirical analysis can be performed in or-
income is often a more useful measure in terms of studies have analyzed European firms, so far. Only studies of Wang (2006) and Goncharov and Hodg- der to assess whether the independent variable
value relevance than comprehensive income. one of the studies has considered IFRS income son (2008) which focus on European firms both also comprehensive income and its separate OCI compo-
As outlined in the previous section, there are only numbers. confirm the stronger predictive power of net income nents can explain the dependent variables future
few empirical studies on the predictive power of over comprehensive income. While Wang (2006) period’s operating cash flows and earnings (link 4).
focuses on local GAAP numbers only, Goncharov If this link between the two defined operational de-
Table 2: and Hodgson (2008) replicate their main tests by finitions is confirmed by empirical evidence, then
Previous studies on the predictive power of comprehensive income and its individual differentiating between local GAAP, IFRS and US inferences about the usefulness of the different in-
components GAAP. They find that under all accounting frame- come numbers for predicting future period’s operat-
Study (year of publication) Country Sample period Conclusions works net income is a better predictor of future op- ing cash flows and earnings can be drawn. In other
Net income predicts future operating cash flows and income erating cash flows than comprehensive income cal- words, the relation between accounting information
Dhaliwal, Subramanyan, and Trezevant (1999) USA 1994 - 1995
better than comprehensive income. culated from all accounting bases. and future firm operating performance becomes
Biddle and Choi (2006) USA 1994 - 1998
No income definition clearly dominates in decision usefulness In summary, it becomes clear that the examination directly observable.
for the prediction of future operating income.
of predictive power needs further clarification. In Note that the predictability could also be analyzed by
Net income is a better predictor of future net
particular, research on the predictive power of IFRS examining the forecast accuracy of analysts before
Kanagaretnam, Mathieu, and Shehata (2009) Canada 1998 - 2003 income/comprehensive income/cash flows from operations
than comprehensive income. income numbers in a homogeneous institutional and after the introduction of IAS 1.81 (rev. 2007).
Comprehensive income is incrementally useful in predicting setting is necessary. To control for differences in However, besides the fact that analysts are likely to
Choi and Zang (2006) USA 1998 - 2003
subsequent period changes in net income. institutional factors (e. g. regulatory requirements use information from multiple sources, no data was
14
Net income always outperforms clean surplus income in
and the enforcement system) is important due to the available to conduct such as study, yet.
Wang (2006) European 1993 - 2002
countries
predicting future firm performance, except in Belgium. fact that otherwise results might be biased by coun- In addition to the assumed validity of links 2 and 3,
try-specific properties. Thus, the current study ad- other factors might also influence the dependent
Comprehensive income can predict subsequent period net
Choi, Das, and Zang (2007) USA 1994 - 2003
income, over and above current period income. dresses the existing research gap by focusing on the variables as defined in operational definition B. For
16 homogeneous institutional background of German this reason, moderating variables are considered.
Net income dominates aggregated comprehensive income in
Goncharov and Hodgson (2008) European 1991 - 2005
predicting cash flows.
IFRS firms. The moderating variables used in this study com-
countries
prise dividends (DIV), a book-to-market ratio
(BTM), a debt-to-equity ratio (D/E), and a dummy

5 6

variable for comprehensive income > net income caused by covenants, etc. (e. g. Dhaliwal, Subrama- H4: Each individual component of other compre- Model 2 (M2):
(CIN). Dividends are included as a control variable nyam, and Trezevant [1999]). Finally, a dummy hensive income has incremental predictive CFOt+1 = α0 + α1*(DNeg_CI) + α2*CIt + α3*(DNeg_CI*CIt)
due to the prior empirical evidence that they contain variable for comprehensive income > net income power for future operating cash flows. + α4…7*moderating variables t + εt
information about expected earnings because firms shall control for manager’s discretion to choose the
target dividends to the permanent component of timing of recognition of components of comprehen- H5: Each individual component of other compre- where CFOt+1 is the operating cash flow in the period
earnings (Miller and Modigliani [1961]). Since the sive income depending upon the underlying eco- hensive income has incremental predictive t+1, NIt is the current (period = t) net income as
market value of a firm represents the current value nomic performance of the firm. According to Choi power for future net income. reported in the income statement and CIt represents
of estimated future cash flows, we use a book-to- and Zhang (2006) managers delay the recognition of current net income adjusted for the components
market ratio to control for any variation in expected unrecognized OCI losses until they become gains or H6: Each individual component of other compre- referred to as current other comprehensive income
earnings missed by the dividend variable (Fama and until the amount of losses reduces in order to avoid hensive income has incremental predictive and as reported in the statement of changes in equi-
French [2000]). A debt-to-equity ratio is included to the negative impact of the loss recognition on net power for future comprehensive income. ty. DNeg_NI and DNeg_CI are dummy variables taking
control for the financial leverage of the firm which income. This may result in higher predictability of the value “1” when NIt is negative or CIt is negative,
might affect future operating performance through positive OCI items for future earnings than negative Finally, we assume that the recent developments in respectively, and “0” otherwise. The use of the
the level of cost of capital, investment restrictions OCI items. the international accounting standards have led to dummy variable and the multiplicative interaction
changes in the usefulness of the different income term (e. g. Braumoeller [2004]) controls for the
Figure 1: numbers for predicting future period’s operating timelier recognition of losses in comparison to gains
Conceptual network of the current study cash flows and earnings. Therefore, the question which is often referred to as conditional conservat-
remains whether the predictive power of income ism (Basu [1997]). As a result, the accuracy of our
numbers has increased or decreased as a result of estimation is improved. The moderating variables
the IASB’s initiatives and actions. As the IASB re- were described in section 4.1. All monetary variables
duces the allowable alternative accounting methods are standardized with total assets at the beginning of
and choices and provides a more consistent ap- the year to avoid that results are primarily driven by
proach to accounting measurement for the goal of larger firms (biased by scale differences).
developing a single set of high quality international Due to the fact that the comparison of adjusted R2s
accounting standards, we predict that these changes of nonnested models is associated with several eco-
in recent years have improved the predictive power nometric problems (e. g. Greene [2003], p. 152-159,
of income numbers (Paananen and Lin [2009]). and Kennedy [2008], p. 87-88) a Vuong (1989) test
Thus, we formulate the following last hypothesis: is performed to investigate whether the difference in
adjusted R2s is significant at conventional levels.
H7: The predictive power of income numbers has H2 is tested by regressing future period’s net income
increased as a consequence of the IASB’s initia- against current period’s net income plus moderating
4.2 Hypotheses tives and actions. variables (Model 3) and against current period’s
In order to provide a starting point for an empirical H2: Comprehensive income has a superior predic- comprehensive income plus moderating variables
analysis the relation among independent and de- tive power for future net income than net in- Having formulated the hypotheses to be tested in (Model 4), respectively. Again, the adjusted R2s of
pendent variables (described as link 4 in Figure 1) is come. the empirical analysis, the following chapter will the models are compared and the difference in coef-
concretized in hypotheses. As the Financial State- present the research design of this study. ficients belonging to the different regressions inves-
ment Presentation Project “shift[s] focus from net H3: Comprehensive income has a superior predic- tigated using the Vuong test. The following two
income to total comprehensive income” (McClain tive power for future comprehensive income 5. Research design cross-sectional models are estimated:
and McLelland [2008], p. 64) it is hypothesized that than net income.
this is done due to the superiority of comprehensive 5.1 Methodology Model 3 (M3):
income over net income in predicting the future In order to also allow for drawing more specific infe- In order to test the hypotheses defined in the pre- NIt+1 = α0 + α1*(DNeg_NI) + α2*NIt + α3*(DNeg_NI*NIt)
operating performance of a firm. Accordingly, we rences about the relevance of individual OCI com- vious section, regression analyses are performed. + α4…7*moderating variables t + εt
state the following first three hypotheses: ponents the following second three hypotheses are For testing hypotheses H1 the adjusted R2s of the
dealing with each OCI component’s predictive pow- following two cross-sectional models (Model 1 and Model 4 (M4):
H1: Comprehensive income has a superior predic- er. In turn, the individual OCI components are ana- Model 2) are compared: NIt+1 = α0 + α1*(DNeg_CI) + α2*CIt + α3*(DNeg_CI*CIt)
tive power for future operating cash flows than lyzed with regard to their incremental predictive + α4…7*moderating variables t + εt
net income. power for future operating cash flows, future net Model 1 (M1):
income and future comprehensive income: CFOt+1 = α0 +α1*(DNeg_NI) + α2*NIt + α3*(DNeg_NI*NIt) where NIt+1 is the net income in the period t+1 while
+ α4…7*moderating variables t + εt NIt, CIt, DNeg_NI, DNeg_CI, and the moderating va-
riables are defined as aforementioned. Similarly, all

7 8
monetary variables are deflated by beginning of the Model 11 (M11): Model 14 (M14)/Model 19 (M19)/Model 24 (M24): residual dispersion measure is calculated for each
year total assets. CIt+1 = α0+ α1*NIt + α2*OCIt + α3…6*moderating CFOt+1/NIt+1/CIt+1 = year of the 10-year time sample, instead. The resi-
The testing of H3 is conducted correspondingly by variables t + εt α0 + α1*(DNeg_CI-ADJ) + α2*CIFOREX-ADJ,t + dual dispersion measure (σε2) is defined as:
estimating the following two cross-sectional models α3*(DNeg_CI-ADJ*CIFOREX-ADJ,t) + α4…7* mod-
(Model 5 and Model 6): Model 12 (M12): erating variables t + εt σε2 = Σ(εi2/[n-k])
CIt+1 = α0+ α1*NIt + α2*AFSt + α3*FOREXt +
Model 5 (M5): α4*HEDGEt + α5*PENSt + α6*REVt + Model 15 (M15)/Model 20 (M20)/Model 25 (M25): where εi is the error term of the regression for each
CIt+1 = α0 + α1*(DNeg_NI) + α2*NIt + α3*(DNeg_NI*NIt) α7…10*moderating variables t + εt CFOt+1/NIt+1/CIt+1 = observation, n is the number of observations of the
+ α4…7*moderating variables t + εt α0 + α1*(DNeg_CI-ADJ) + α2*CIHEDGE-ADJ,t + year and k is the number of independent variables
where CFOt+1, NIt+1, CIt+1, NIt, CIt, DNeg_NI, DNeg_CI, α3*(DNeg_CI-ADJ*CIHEDGE-ADJ,t) + α4…7* mod- including the intercept term.
Model 6 (M6): and the moderating variables are defined as afore- erating variables t + εt Next, we test for a general decrease of the residual
CIt+1 = α0 + α1*(DNeg_CI) + α2*CIt + α3*(DNeg_CI*CIt) mentioned. AFSt is the current unrealized gain or dispersion measure (increase of predictive power)
+ α4…7*moderating variables t + εt loss on re-measuring available-for-sale financial Model 16 (M16)/Model 21 (M21)/Model 26 (M26): over the entire sample period in regressions on time
assets, FOREXt is the current foreign currency trans- CFOt+1/NIt+1/CIt+1 = metrics. For that purpose we estimate the following
where CIt+1 is the comprehensive income in the pe- lation adjustment, HEDGEt is the current gain or α0 + α1*(DNeg_CI-ADJ) + α2*CIPENS-ADJ,t + raw and rank regression models (Model 28 and
riod t+1 while again NIt, CIt, DNeg_NI, DNeg_CI, and the loss on cash flow hedges or on hedging instruments α3*(DNeg_CI-ADJ*CIPENS-ADJ,t) + α4…7* mod- Model 29):
moderating variables are defined as aforementioned. of a net investment in a foreign operation, PENSt is erating variables t + εt
Likewise, all monetary variables are scaled by total the current actuarial gain or loss on defined benefit Model 28 (M28):
assets as of beginning of the year. By the same token, pension obligations, and REVt is the current unrea- Model 17 (M17)/Model 22 (M22)/Model 27 (M27): σε2t = β0+ β1*t + γt
the adjusted R2s of the two models are compared lized gain or loss due to the revaluation of intangible CFOt+1/NIt+1/CIt+1 =
and the difference investigated for significance using assets and/or property, plant and equipment. In α0 + α1*(DNeg_CI-ADJ) + α2*CIREV-ADJ,t + Model 29 (M29):
the Vuong test. conformity with the models above, all monetary α3*(DNeg_CI-ADJ*CIREV-ADJ,t) + α4…7* mod- Rank_σε2t = β0+ β1*t + γt
For testing H4 through H6 the basic models pre- variables are scaled by beginning of the year total erating variables t + εt
sented above are extended and include net income, assets. where σε2t is the residual dispersion measure of the
aggregated other comprehensive income and/or the The resulting adjusted R2s from the regressions where CFOt+1, NIt+1, CIt+1, and the moderating va- regressions for period t, Rank_σε2t is the rank of the
individual other comprehensive components as the above are then compared to the adjusted R2 of M1, riables are defined as aforementioned. CIAFS-ADJ,t is residual dispersion measure for the whole time pat-
main independent variables. The dependent va- M3, and M5, respectively. Again, differences in coef- current net income adjusted for AFSt, CIFOREX-ADJ,t is tern (e. g. “1” for the smallest measure and “10” for
riables are again derived to be future operating cash ficients are tested for significance using Vuong’s current net income adjusted for FOREXt, CIHEDGE- the largest measure), and t is the sample year
flows, future net income and future comprehensive diagnostic statistic. This is done in order to test ADJ,t is current net income adjusted for HEDGEt, (t = 1…10 corresponds to years 1998…2007).
income, respectively. The estimated cross-sectional whether the dirty surplus components (individually CIPENS-ADJ,t is current net income adjusted for PENSt, While M28 imposes a linear relation of the residual
models are presented below (Model 7 through to or in the aggregate) have incremental value that and CIREV-ADJ,t is current net income adjusted for dispersion metric to time, M29 imposes no particu-
Model 12): increases the predictability of future firm operating REVt. DNeg_CI-ADJ is a dummy variable taking the lar functional form. As we have no conceptual or
performance. value “1“ when CIt adjusted for the respective indi- other basis for favoring one functional form over
Model 7 (M7): Subsequently, in order to evaluate the appropriate- vidual component is negative and “0” otherwise. another, the rank regression approach provides a
CFOt+1 = α0+ α1*NIt + α2*OCIt + α3…6*moderating ness of current individual components of other Differences in the adjusted R2s of M13-M27 are also reasonable compromise between power and the risk
variables t + εt comprehensive income more specifically, we ex- tested for significance against M1, M3, and M5, re- of imposing a misspecified functional form (Francis
amine all individual components on an item-by-item spectively, using Vuong’s diagnostic statistic. and Schipper [1999], pp. 327-328). In fact, M29
Model 8 (M8): basis. This is done by adjusting net income indivi- Finally, the analyses of the incremental value of dirty greatly serves as a backup for the raw regression
CFOt+1 = α0+ α1*NIt + α2*AFSt + α3*FOREXt dually for each component of other comprehensive surplus components for the prediction of future firm results of M28.
+ α4*HEDGEt + α5*PENSt + α6*REVt income and then comparing the association between operating performance are expanded to include
+ α7…10*moderating variables t + εt each of the resulting measures of adjusted income predictions of future operating cash flows, future net 5.2 Sample
and future firm operating performance with the income and future comprehensive income beyond The analyses use data from German companies
Model 9 (M9): results of the basic models. The following cross- one period. More specifically, the cross-sectional belonging to the German major stock exchange in-
NIt+1 = α0+ α1*NIt + α2*OCIt + α3…6*moderating sectional regression models are estimated (Model 13 regressions include lagged independent variables of dex ‘HDAX’ as of August 16, 2009. All accounting
variables t + εt through to Model 27): the period t-1, t-2, t-3, and t-4. According to Ber- data had to be hand-collected due to the fact that
nard (1995), p. 739: “There is evidently little to be common databases such as Worldscope and Hop-
Model 10 (M10): Model 13 (M13)/Model 18 (M18)/Model 23 (M23): gained by forecasting earnings […] beyond four penstedt lack detail and quality with regard to (oth-
NIt+1 = α0+ α1*NIt + α2*AFSt + α3*FOREXt CFOt+1/NIt+1/CIt+1 = years!”. er) comprehensive income reporting. Only, the mar-
+ α4*HEDGEt + α5*PENSt + α6*REVt α0 + α1*(DNeg_CI-ADJ) + α2*CIAFS-ADJ,t + Next, H7 is tested by segmenting the regression ket value of common equity used to calculate the
+ α7…10*moderating variables t + εt α3*(DNeg_CI-ADJ*CIAFS-ADJ,t) + α4…7* mod- results by time period. Due to the fact that adjusted book-to-market ratio is obtained from Thomson
erating variables t + εt R2s are incomparable across samples, Gu’s (2007) Financial Analytics – Worldscope.

9 10

The HDAX index includes the largest (in terms of panies or 370 firm-year observations. We test for the
market capitalization) and most actively traded capi- robustness of results of the reduced sample by re- Table 3:
tal market-oriented German firms which are closely running all regressions without moderating va- Descriptive statistics
monitored by analysts and often serve as a bench- riables in section 6.6. To reduce the bias created by
Standard
mark for many smaller firms with regard to the ap- outliers, the top and bottom 1% of the data is winso- Mean Median Minimum Maximum
deviation
N
plication of IFRS accounting. rized (Field [2005], p. 74 cont.).
CFO 0.09963 0.08752 -0.27366 0.56684 0.11481 463
The focus on one capital market allows controlling
for institutional factors such as regulatory and listing NI 0.06169 0.04952 -0.23243 0.36978 0.08761 463
6. Empirical results
requirements or the enforcement system CI 0.06002 0.04982 -0.24438 0.38787 0.08962 463

(Ernstberger [2008], p. 14). In addition, Germany is 6.1 Descriptive statistics OCI -0.00193 -0.00030 -0.05391 0.05473 0.01608 463
a country where a sufficient sample size for IFRS The descriptive statistics of the deflated variables AFS 0.00029 0.00000 -0.01487 0.02466 0.00417 463
accounting data is available beginning from the late used in the analyses are presented in Table 3. As can FOREX -0.00254 -0.00052 -0.03766 0.04017 0.01190 463
1990s. Since 1998 many listed companies have exer- be seen, the overview documents that the majority of HEDGE 0.00008 0.00000 -0.02591 0.03095 0.00619 463
cised an option provided by law (Kapitalaufnah- companies have positive operating cash flows, a PENS 0.00013 0.00000 -0.01976 0.01405 0.00384 463
meerleichterungsgesetz – Capital Raising Act) to positive net income, and a positive comprehensive REV 0.00011 0.00000 -0.00090 0.00810 0.00089 463
prepare consolidated financial statements according income. In contrast, other comprehensive income is DIV 0.00630 0.00000 0.00000 0.10935 0.01441 463
to IFRS. mainly negative which causes comprehensive in- BTM 0.62570 0.46781 0.04633 2.70130 0.49379 370
The initial sample consists of the 12-month period come to be lower than net income on average. This is D/E 2.63250 1.83840 0.10373 27.16012 3.69060 463
firm-year observations between 1998 and 2007 from also documented by a mean value of 0.42333 for the CIN 0.42333 0.00000 0.00000 1.00000 0.49462 463
all 110 companies which were listed in the major dummy variable CIN. Correspondingly to the results
segments of the German Stock Exchange. In order to of Ernstberger [2008] and in contrast to prior stu- Notes: The panel shows the descriptive statistics for the entire sample of IFRS com-
be able to include reporting periods which do not panies belonging to the German major stock exchange index ‘HDAX’ as of August 16,
dies, the mean and median of FOREX mostly have a 2009. Observations for which the absolute value of the test variable falls in the top or
equal the calendar year, ‘year’ was defined as the negative sign and drive the results of other compre- bottom percentile of the test-variable distribution are winsorized. CFO is the operat-
year in which the fiscal period ends when the end of hensive income. This might be due to the fact that ing cash flows, NI is the net income, CI is the comprehensive income, OCI is the
the fiscal period is between and including June 30 many German companies have subsidiaries in other other comprehensive income, AFS is the unrealized gain or loss on re-measuring
and December 31 and the previous year otherwise. available-for-sale financial assets, FOREX is the foreign currency translation adjust-
countries and the Euro appreciated against other ment, HEDGE is the gain or loss on cash flow hedges or on hedging instruments of a
All companies under examination had to apply full major currencies in the period examined. net investment in a foreign operation, PENS is the actuarial gain or loss on defined
IFRS accounting in a period and not only provide The mean of AFS, HEDGE, PENS, and REV is posi- benefit pension obligations, REV is the unrealized gain or loss due to the revaluation
reconciliations. In accordance with previous studies, tive while the median, respectively, is equal to “0”. of intangible assets and/or property, plant and equipment, DIV is the dividends,
we eliminate banks, insurance companies and other BTM is the book-to-market ratio, D/E is the debt-to-equity ratio, and CIN is a dum-
This is consistent with prior studies for IFRS OCI my variable that takes the value “1” when comprehensive income > net income and
financial firms from the sample reducing our sample data (Ernstberger [2008]) as well as US OCI data “0” otherwise, respectively, of the period. N is the number of firm-year observations
size to 98 companies. This is done due to the fact (e. g. Kanagaretnam, Mathieu, and Shehata [2009]; between 1998 and 2007. All variables are scaled by total assets as of beginning of the
that the balance sheet structures of these firms are Dhaliwal, Subramanyan, and Trezevant [1999]). year.
fundamentally different to those of non-financial Reasons for the positive mean values might be the
firms and would not allow for comparison. Finally, increase in stock prices and for the median values of 6.3 Regression results for basic models time Vuong’s (ν) diagnostic statistic confirms that
we exclude all observations representing a short “0” that many companies do not report such items. The adjusted R2s from estimating M1 through to M6 the difference is statistically significant at the 0.01
financial year. Further, the BTM mean value of 0.62750 shows no are presented in Table 4, panel A to panel C. As can two-tailed level. Therefore, H2 also has to be re-
The final sample consists of 96 companies or 463 signs of overvaluation, on average. Finally, the D/E be seen in panel A of Table 4, we find no superior jected.
firm-year observations. Unfortunately, some miss- mean of 2.63250 indicates that companies are main- predictive power of comprehensive income for fu- Table 4, panel C shows the results of the examina-
ing values of the market value of common equity ly financed by debt in order to reduce their cost of ture operating cash flows than net income. In fact, tion of the association of net income vs. comprehen-
further reduces our sample size for the regression capital (e. g. D’Mello and Farhat [2008]). the adjusted R2 of 0.1158 for M1 is higher than the sive income with future comprehensive income. In
analyses including moderating variables to 71 com- adjusted R2 of 0.1098 for M2. However, the Vuong conformance with the results above, the adjusted R2
test indicates that the difference in adjusted R2s is of 0.4395 for the net income model (M5) is higher
not significant at conventional levels. Nevertheless, than the adjusted R2 of 0.4238 for the comprehen-
H1 has to be rejected as a consequence of the results. sive income model (M6). The ν-statistic indicates
Table 4, panel B reports the results of the estimation that the difference is significant at the 4%-level.
of models that test the association of net income vs. Thus, we find no superior predictive power of com-
comprehensive income with future net income. prehensive income for future comprehensive income
Again, the adjusted R2 of 0.4550 for the net income than net income. H3 also has to be rejected.
model (M3) is higher than the adjusted R2 of 0.4364 In line with prior studies, the adjusted R2s for the
for the comprehensive income model (M4). This models using future operating cash flows as the

11 12
independent variable are higher than for the models 6.4 Regression results for extended models tional levels. Thus, we find no evidence for the claim Table 6:
using net income or comprehensive income as the The results of estimating M7 through to M12 are that each individual component of OCI has (signifi- Results from estimating Model 13 through to
independent variable. These results indicate a better reported in Table 5, panel A to panel C. The adjusted cant) incremental predictive power for future oper- Model 27
overall predictability of future net income and future R2s of these models are compared to the basic mod- ating cash flows (rejection of H4).
Panel A: Association of net income adjusted for individual
comprehensive income in comparison to future els (e. g. M1, M3, or M5) in order to test whether the components of OCI with future operating cash flows
operating cash flows. OCI components (individually or in the aggregate) Table 5: Model
a
Adj. R
2 ,b
ν-statistic
c
ρ-value
d
N
e

have incremental value that enhances the predicta- Results from estimating Model 7 through to M13 0.1140 0.9934 0.32 370
Table 4: bility of future firm operating performance. This Model 12 M14 0.1153 0.0018 1.00 370
Results from estimating Model 1 through to examination marks a first step towards testing H4 M15 0.1144 0.5630 0.57 370
Panel A: Association of net income and (individual components
Model 6 through to H6. of) OCI with future operating cash flows M16 0.1149 0.5628 0.57 370
The regression results in panel A of 0.1256 for M7 Model
a
Adj. R
2 ,b
ν-statistic
c
ρ-value
d
N
e
Panel A: Association of net income vs. comprehensive income M17 0.1154 -1.0255 0.31 370
with future operating cash flows and of 0.1218 for M8 suggest that the inclusion of M7 0.1256 -1.0120 0.31 370
Model
a
Adj. R
2 ,b
ν-statistic
c
ρ-value
d
N
e
both aggregated and individual OCI components M8 0.1218 -1.4518 0.15 370
Panel B: Association of net income adjusted for individual
components of OCI with future net income
M1 0.1153 370 improve the predictability of future operating cash a 2 ,b c d e
Panel B: Association of net income and (individual components Model Adj. R ν-statistic ρ-value N
M2 0.1098 1.3023 0.19 370 flows as these adjusted R2s are higher than the ad-
of) OCI with future net income M18 0.4536 0.7769 0.44 370
justed R2 of 0.1153 for the basic model M1. Mean- Model
a
Adj. R
2 ,b
ν-statistic
c
ρ-value
d
N
e
Panel B: Association of net income vs. comprehensive income M19 0.4546 0.0807 0.94 370
while, the ν-statistics show that the differences in M9 0.4583 -0.6820 0.50 370
with future net income
a 2 ,b c d e adjusted R2s are not statistically significant. Panel B M20 0.4486 1.7551 0.08 370
Model Adj. R ν-statistic ρ-value N M10 0.4651 -1.4530 0.15 370
M3 0.4550 370
of Table 5 confirms these results with regard to pre- M21 0.4499 2.4653 0.01 370

dicting future net income. Though the adjusted R2s Panel C: Association of net income and (individual components M22 0.4549 0.4224 0.67 370
M4 0.4364 2.4852 0.01 370 of) OCI with future comprehensive income
of 0.4583 for M9 and of 0.4651 for M10 are higher a 2 ,b c d e
Panel C: Association of net income adjusted for individual
Model Adj. R ν-statistic ρ-value N
Panel C: Association of net income vs. comprehensive income than the adjusted R2 of 0.4550 for M3, the Vuong components of OCI with future comprehensive income
with future comprehensive income M11 0.4402 -0.4450 0.66 370 a 2 ,b c d e
a 2 ,b c d e diagnostic statistic indicates that the results are not Model Adj. R ν-statistic ρ-value N
Model Adj. R ν-statistic ρ-value N M12 0.4472 -1.2975 0.19 370
significant at conventional levels. The adjusted R2s M23 0.4382 0.6594 0.51 370
M5 0.4395 370
for the association of net income and (individual Notes: The panels show the results of the estimation of models M24 0.4401 -0.1268 0.90 370
M6 0.4238 2.0459 0.04 370
components of) OCI are reported in panel C of Ta- that test the association of net income and (individual components M25 0.4325 1.7516 0.08 370
ble 5. As can be seen there, the adjusted R2s of of) OCI with future operating cash flows (panel A), with future net
Notes: The panels show the results of the estimation of models M26 0.4352 2.0485 0.04 370
income (panel B), and with future comprehensive income (panel
that test the association of net income and comprehensive income 0.4402 for M11 and of 0.4651 for M12 are higher
C). The sample used for the cross-sectional regressions consists of M27 0.4394 0.1125 0.91 370
with future operating cash flows (panel A), with future net income than the adjusted R2 of 0.4395 for the basic model all IFRS companies belonging to the German major stock ex-
(panel B), and with future comprehensive income (panel C). The
M5. However, the ν-statistics again present that the change index ‘HDAX’ as of August 16, 2009. Notes: The panels show the results of the estimation of models
sample used for the cross-sectional regressions consists of all IFRS
differences in adjusted R2s are not significant at aModel M7: CFOt+1 = α0+ α1*NIt + α2*OCIt + α3…6*moderating that test the association of net income adjusted for individual
companies belonging to the German major stock exchange index
variables t + εt components of OCI with future operating cash flows (panel A),
‘HDAX’ as of August 16, 2009. conventional levels. Overall, the insignificant results with future net income (panel B), and with future comprehensive
aModel M1: CFOt+1 = α0 +α1*(DNeg_NI) + α2*NIt + α3*(DNeg_NI*NIt)
Model M8: CFOt+1 = α0+ α1*NIt + α2*AFSt + α3*FOREXt
support a rejection of H4 through to H6. + α4*HEDGEt + α5*PENSt + α6*REVt income (panel C). The sample used for the cross-sectional regres-
+ α4…7*moderating variables t + εt
Next, to evaluate the appropriateness of current + α7…10*moderating variables t + εt sions consists of all IFRS companies belonging to the German
Model M2: CFOt+1 = α0 + α1*(DNeg_CI) + α2*CIt + α3*(DNeg_CI*CIt)
individual components of OCI more specifically M13 Model M9: NIt+1 = α0+ α1*NIt + α2*OCIt + α3…6*moderating va- major stock exchange index ‘HDAX’ as of August 16, 2009.
+α4…7*moderating variables t + εt aModel M13: CFOt+1 =
riables t + εt α0 + α1*(DNeg_CI-ADJ) + α2*CIAFS-ADJ,t
Model M3: NIt+1 = α0 +α1*(DNeg_NI) + α2*NIt + α3*(DNeg_NI*NIt) through to M27 are estimated. As described above, + α3*(DNeg_CI-ADJ*CIAFS-ADJ,t) + α4…7* mod-
Model M10: NIt+1 = α0+ α1*NIt + α2*AFSt + α3*FOREXt
+ α4…7*moderating variables t + εt net income is adjusted individually for each OCI + α4*HEDGEt + α5*PENSt + α6*REVt erating variables t + εt
Model M4: NIt+1 = α0 + α1*(DNeg_CI) + α2*CIt + α3*(DNeg_CI*CIt)
component and then tested for its association with + α7…10*moderating variables t + εt Model M14: CFOt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIFOREX-ADJ,t
+α4…7*moderating variables t + εt
future firm operating performance. The regression Model M11: CIt+1 = α0+ α1*NIt + α2*OCIt + α3…6*moderating va- + α3*(DNeg_CI-ADJ*CIFOREX-ADJ,t) + α4…7*
Model M5: CIt+1 = α0 +α1*(DNeg_NI) + α2*NIt + α3*(DNeg_NI*NIt)
riables t + εt moderating variables t + εt
+ α4…7*moderating variables t + εt results are then compared to the basic models M1, Model M15: CFOt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIHEDGE-ADJ,t
Model M12: CIt+1 = α0+ α1*NIt + α2*AFSt + α3*FOREXt
Model M6: CIt+1 = α0 + α1*(DNeg_CI) + α2*CIt + α3*(DNeg_CI*CIt) M3, or M5 as reported in Table 6 (panel A to pan- + α4*HEDGEt + α5*PENSt + α6*REVt + α3*(DNeg_CI-ADJ*CIHEDGE-ADJ,t) + α4…7*
+α4…7*moderating variables t + εt
bAdj. R2 reports the estimated adjusted coefficient of determina-
el C). + α7…10*moderating variables t + εt moderating variables t + εt
Panel A of Table 6 presents the results for the asso- bAdj. R2 reports the estimated adjusted coefficient of determina- Model M16: CFOt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIPENS-ADJ,t
tion on the relevant model.
tion on the relevant model. + α3*(DNeg_CI-ADJ*CIPENS-ADJ,t) + α4…7*
cAdj. R2s are compared using the likelihood ratio test described in ciation of net income adjusted for individual com- cAdj. R2s are compared using the likelihood ratio test described in moderating variables t + εt
Vuong (1989). The ν-statistic returns the value of Vuong’s diagnos- ponents of OCI with future operating cash flows. As Vuong (1989). The ν-statistic returns the value of Vuong’s diagnos- Model M17: CFOt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIREV-ADJ,t
tic statistic comparing the explanatory power of a model with the
can be seen, the adjusted R2s are lower for the AFS-, tic statistic comparing the explanatory power of a model with the + α3*(DNeg_CI-ADJ*CIREV-ADJ,t) + α4…7* mod-
explanatory power of the basic model (e. g. M1, M3, or M5).
FOREX-, HEDGE-, and PENS-adjustment, respec- explanatory power of the basic model (e. g. M1, M3, or M5). erating variables t + εt
dρ-value indicates the two-tailed significance level for the differ-
dρ-value indicates the two-tailed significance level for the differ- Model M18: NIt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIAFS-ADJ,t
ence in the explanatory power of a model and the explanatory tively, and only higher for the REV-adjustment in + α3*(DNeg_CI-ADJ*CIAFS-ADJ,t) + α4…7* mod-
ence in the explanatory power of a model and the explanatory
power of the basic model based on Vuong’s likelihood ratio test. comparison to M1. As indicated by the ν-statistics, power of the basic model based on Vuong’s likelihood ratio test. erating variables t + εt
eN is the number of firm-year observations between 1998 and
none of the differences between the adjusted R2s of eN is the number of firm-year observations between 1998 and
2007.
M13 through to M17 and M1 is significant at conven- 2007.

13 14

Table 6 (Notes continued) be rejected as a consequence of the estimation re- Table 7: Table 7 (Notes continued)
Model M19: NIt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIFOREX-ADJ,t sults. Optimized prediction model of future firm bThe dependent variables CFOt+1, NIt+1, and CIt+1, respectively, are

+ α3*(DNeg_CI-ADJ*CIFOREX-ADJ,t) + α4…7* mod- defined as aforementioned (section 5.1).


erating variables t + εt
Table 6, panel C reports the regression results for operating performance cAdj. R2 reports the estimated adjusted coefficient of determina-

Model M20: NIt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIHEDGE-ADJ,t the association of net income adjusted for individual b tion on the relevant model.
Independent Dependent variable:
+ α3*(DNeg_CI-ADJ*CIHEDGE-ADJ,t) + α4…7* mod- components of OCI with comprehensive income variable:
a
CFOt +1 ρ-value e
NI t +1 ρ-valuee CI t +1 ρ-value
e dAdj. R2s are compared using the likelihood ratio test described in

erating variables t + εt (M23 through to M27). Similar to the results of pan- Vuong (1989). The ν-statistic returns the value of Vuong’s diagnos-
Intercept 0.0468 0.02 0.0195 0.24 0.0221 0.17
Model M21: NIt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIPENS-ADJ,t tic statistic comparing the explanatory power of a model with the
+ α3*(DNeg_CI-ADJ*CIPENS-ADJ,t) + α4…7* mod-
el B, the adjusted R2s of the regression models are NI t 0.3487 0.00 0.6449 0.00 0.7570 0.00 explanatory power of the basic model (e. g. M1, M3, or M5).
erating variables t + εt all not significant at conventional levels except for dρ-value indicates the one-tailed significance level of the t-statistic
NI t -1 0.1234 0.28 0.1930 0.11
Model M22: NIt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIREV-ADJ,t the PENS-adjustment (M26) which is significant at on the relevant variable, or respectively, the two-tailed significance
+ α3*(DNeg_CI-ADJ*CIREV-ADJ,t) + α4…7* mod- NI t -2 -0.2715 0.01 -0.2745 0.00 level for the difference in the explanatory power of a model and the
the 4%-level. Thus, the problem of excluding some
erating variables t + εt NI t -3 -0.3229 0.02 explanatory power of the basic model based on Vuong’s likelihood
Model M23: CIt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIAFS-ADJ,t
future events from the estimation process of actuari- ratio test.
+ α3*(DNeg_CI-ADJ*CIAFS-ADJ,t) + α4…7* mod- al gains and losses on defined benefit pension obli- NI t -4 0.2637 0.06 0.1845 0.10

erating variables t + εt gations can also be confirmed for the comprehensive AFSt 2.3118 0.08 2.0762 0.10
regard to future operating cash flows; for AFS, FO-
Model M24: CIt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIFOREX-ADJ,t level of future income. In conclusion, H6 has to be
+ α3*(DNeg_CI-ADJ*CIFOREX-ADJ,t) + α4…7* mod-
AFSt -2 -2.2757 0.16
REX, HEDGE, PENS, and REV with regard to future
erating variables t + εt
rejected accordingly. FOREXt -0.8183 0.09 net income; and for AFS, FOREX, HEDGE, and
Model M25: CIt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIHEDGE-ADJ,t Finally, after showing that not any individual com- FOREXt -1 -0.6646 0.08 PENS with regard to future comprehensive income.
+ α3*(DNeg_CI-ADJ*CIHEDGE-ADJ,t) + α4…7* mod- ponent of OCI has (significant) incremental predic- This adds empirical evidence to the theoretical
FOREXt -3 -0.9675 0.03
erating variables t + εt tive power for subsequent period’s firm operating
Model M26: CIt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIPENS-ADJ,t HEDGEt -2.2728 0.01 -2.2235 0.01 -1.8750 0.01
framework of Ohlson (1999) which concludes “that
+ α3*(DNeg_CI-ADJ*CIPENS-ADJ,t) + α4…7* mod-
performance, we expand our analyses of M8, M10, transitory earnings have no effect on next-period
HEDGEt -4 2.1802 0.10 1.5499 0.22
erating variables t + εt and M12 to include lagged independent variables of expected earnings if and only if they are ‘passed on’
Model M27: CIt+1 = α0 + α1*(DNeg_CI-ADJ) + α2*CIREV-ADJ,t the period t-1, t-2, t-3, and t-4. In other words: we PENSt -1 1.7488 0.14 -2.5559 0.03
as dividends”.
+ α3*(DNeg_CI-ADJ*CIREV-ADJ,t) + α4…7* mod- specifically examine the predictive power of all OCI PENSt -2 3.2961 0.03 4.0164 0.00 2.4182 0.07
erating variables t + εt In summary, the acceptance or rejection of H4
bAdj. R2 reports the estimated adjusted coefficient of determina-
components beyond one period. After estimating the PENSt -4 12.5059 0.18 15.1233 0.14 through to H6 depends on the specific prediction
tion on the relevant model. regression models a high number of coefficients are REVt -12.1391 0.06 period under examination. While we fail to find
cAdj. R2s are compared using the likelihood ratio test described in found to have no statistically significant impact on REVt -4 -8.4132 0.25 significant incremental predictive power of OCI
Vuong (1989). The ν-statistic returns the value of Vuong’s diagnos- the regression. Hence, a step-wise elimination
tic statistic comparing the explanatory power of a model with the DIVt 0.6875 0.14 1.3760 0.00 1.2354 0.00 components for subsequent period’s firm operating
explanatory power of the basic model (e. g. M1, M3, or M5).
process is carried out where the variable with the performance, we are able to confirm the incremental
DIVt -3 -1.3216 0.16
dρ-value indicates the two-tailed significance level for the differ- highest ρ-value is eliminated before the regression predictive power of OCI components beyond one
ence in the explanatory power of a model and the explanatory model is estimated again. In case of an increased DIVt -4 1.0809 0.01 1.3108 0.14
period.
power of the basic model based on Vuong’s likelihood ratio test. explanatory power of the “new” model the left out BTMt -0.0464 0.06 -0.0692 0.00 -0.0916 0.00
eN is the number of firm-year observations between 1998 and

2007.
variable remains eliminated and the procedure is BTMt -1 0.0480 0.04 0.0621 0.00
6.5 Time-series regressions
continued with the next variable until all remaining BTMt -2 0.0455 0.04 Table 8 reports the results for Gu’s (2007) residual
The results of estimating M18 through to M22 are variables of the equation have significant incremen- BTMt -4 -0.0178 0.19 dispersion measure from estimating M1 through to
presented in panel B of Table 4. These models test tal explanatory power. We have not included a table M6. The results are segmented by time period in
D/Et -0.0028 0.07
the association of net income adjusted for individual picturing the elimination process performed due to order to analyze over time changes in the predictive
D/Et -1 -0.0075 0.09
components of OCI with future net income. The space limitation. However, the “statistically opti- power of net income and comprehensive income for
D/Et -3 0.0060 0.10
adjusted R2s of the regression models are all lower mized” (lagged) prediction models for future operat- future firm operating performance. The number of
than the adjusted R2 of 0.4550 for the basic model ing cash flows, future net income and future com- CINt 0.0322 0.03 0.0187 0.14
observations in the annual regressions ranges from
M3. However, the Vuong tests show that all differ- prehensive income are presented in Table 7. CINt -2 0.0360 0.00
10 in 1998 to 71 in 2007. On average, the residual
ences in adjusted R2s are not statistically significant As can be seen there, the adjusted R2s of the opti- 2 ,c
dispersion measure is lowest for predicting future
Adj. R 0.5488 0.7255 0.7343
except for the PENS-adjustment which is significant mized prediction models (0.5488 for CFOt+1, 0.7255 net income, followed by future comprehensive in-
d

at the 0.01 two-tailed level. A reason for the “noise” for NIt+1, and 0.7343 for CIt+1) are higher than for ν-statistic -4.8329 0.00 -6.7562 0.00 -5.5273 0.00
come and future operating cash flows.
added by the PENS-adjustment could be the follow- M8, M10, or M12, respectively. In addition, the dif- Notes: The table shows the results of the estimation of “optimized” Figure 2 displays the over time patterns of the resi-
ing. The actuarial gains and losses on defined benefit ferences in adjusted R2s are significant at conven- (in terms of explanatory power) models that test the association of dual dispersion measures of the relevant models.
tional levels (ν-statistic at lower than the 0.01 two- net income and individual components of OCI with future operat-
pension obligations include the effects of some fu- ing cash flows, with future net income, and with future compre- The graphs show a distinct increase for all models
ture events, while others are excluded. For example, tailed level). Thus, some individual OCI components for the time period under examination. Thus, the
hensive income, respectively, of up to 4-years. The sample used for
mortality and salary progression are estimated, seem to have incremental predictive power beyond the cross-sectional regressions consists of all IFRS companies predictive power of net income and comprehensive
while future plan amendments and future years of net income when longer time periods are consi- belonging to the German major stock exchange index ‘HDAX’ as of for future firm operating performance seems to have
dered. In particular, we find predictive power August 16, 2009.
service are not anticipated regardless of the proba- aAll independent variables in the relevant models are defined as deteriorated as a consequence of the IASB’s recent
bility of occurrence (Beaver [1991]). Again, H5 has to beyond one period for HEDGE, PENS and REV with initiatives and actions (rejection of H7).
aforementioned (section 5.1).

15 16
Table 8: Figure 2:
Contemporaneous relations between earnings and future firm operating performance Residual dispersion measure over time
Model
a 0.016
Year N
b 2 ,c 2 ,c 2 ,c 2 ,c 2 ,c 2 ,c M1 - residual dispersion
M1 - σε M2 - σε M3 - σε M4 - σε M5 - σε M6 - σε
M2 - residual dispersion
1998 10 0.0020 0.0033 0.0002 0.0002 0.0007 0.0008 M3 - residual dispersion
0.014
1999 12 0.0030 0.0020 0.0001 0.0001 0.0001 0.0001 M4 - residual dispersion
M5 - residual dispersion
2000 15 0.0031 0.0030 0.0007 0.0007 0.0011 0.0011
M6 - residual dispersion
2001 25 0.0031 0.0030 0.0011 0.0010 0.0015 0.0014 0.012

2002 31 0.0068 0.0073 0.0052 0.0060 0.0053 0.0058


2003 32 0.0070 0.0065 0.0050 0.0052 0.0044 0.0044 0.01
2004 44 0.0152 0.0147 0.0055 0.0055 0.0056 0.0056
2005 62 0.0146 0.0146 0.0028 0.0029 0.0030 0.0031
0.008
2006 68 0.0121 0.0129 0.0056 0.0061 0.0063 0.0069
2007 71 0.0101 0.0102 0.0053 0.0054 0.0057 0.0058

Average
d
0.0077 0.0078 0.0032 0.0033 0.0033 0.0035 0.006

Notes: The panel shows the results for Gu’s (2007) residual dispersion measure of the
estimation of models that test the association of net income and comprehensive 0.004
income with future operating cash flows (M1 and M2), with future net income (M3
and M4), and with future comprehensive income (M5 and M6). The sample used for
the cross-sectional regressions consists of all IFRS companies belonging to the Ger- 0.002
man major stock exchange index ‘HDAX’ as of August 16, 2009.
aModel M1: CFOt+1 = α0 +α1*(DNeg_NI) + α2*NIt + α3*(DNeg_NI*NIt) + α4…7*moderating

variables t + εt 0
Model M2: CFOt+1 = α0 + α1*(DNeg_CI) + α2*CIt + α3*(DNeg_CI*CIt) +α4…7*moderating 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
variables t + εt
Model M3: NIt+1 = α0 +α1*(DNeg_NI) + α2*NIt + α3*(DNeg_NI*NIt) + α4…7*moderating
variables t + εt
We test the significance of these over time patterns 2003 is the same as the trend in 2004-2007. A sig-
Model M4: NIt+1 = α0 + α1*(DNeg_CI) + α2*CIt + α3*(DNeg_CI*CIt) +α4…7*moderating by estimating raw and rank regressions as defined in nificant structural brake (at less than the 0.05 level)
variables t + εt M28 and M29 above. Turning to the results of the can also be verified for the predictability of future
Model M5: CIt+1 = α0 +α1*(DNeg_NI) + α2*NIt + α3*(DNeg_NI*NIt) + α4…7*moderating regressions (Table 9), we see that the time coeffi- net income at year 2002. Thus, for future net in-
variables t + εt
Model M6: CIt+1 = α0 + α1*(DNeg_CI) + α2*CIt + α3*(DNeg_CI*CIt) +α4…7*moderating
cient β1 is always positive at conventional levels. come the trend in predictive power in 1998-2001 is
variables t + εt Thus, we find a significant increase in the residual different than the trend in 2002-2007. However, we
bN is the number of firm-year observations between 1998 and 2007. dispersions measure over time. This trend is also find no structural brake for the predictability of fu-
cThe residual dispersion measure (σε2) is defined as: σε2 = Σ(εi2/[n-k]), where εi is the
confirmed by the results of the rank regressions. The ture comprehensive income over time.
error term of the regression for each observation, n is the number of observations of
the year and k is the number of independent variables including the intercept term.
estimated magnitude of the decline of predictive These results correspond to the results of Paananen
dThe row labeled average reports the mean of the year residual dispersion estimates. power of net income and comprehensive income for and Lin (2009) who find a worsening of accounting
future firm operating performance (as indicated by quality over time, especially for the period after
the raw regression results) lies in between 18.18 % 2004. Further analysis show that this development
and 21.88 % per year with respect to the average. is less likely to be driven by new adopters of IFRS
Next, Chow (1960) tests on structural change are but is driven by the changes of the standards. The
conducted to see if parameters have changed during structural breaks identified in this study correspond
the time period. Pinpointing the exact timing of a to the shifts from the IAS period to the IFRS volun-
structural change is a difficult task and tests of this tary period in 2002 and from the IFRS voluntary
nature typically have low powers (Greene, 2003 period to the IFRS mandatory period in 2004.
Ch. 7). As visually indicated by Figure 2, a structural Please note that we have tested the sample for hete-
brake could be possible at year 2004 for future op- roscedasticity using White’s (1980) test and for au-
erating income, and at year 2002 for future net in- tocorrelation using the Durbin-Watson (1950/1951)
come and future comprehensive income, respective- test as well as the Breusch-Godfrey LM (1978) test.
ly. The Chow test reliably rejects (at less than the The null hypotheses of “heteroscedasticity not
0.01 level) the null hypothesis that the trend for the present” and “no autocorrelation”, respectively,
predictability of future operating cash flows in 1998- could not be rejected at conventional levels.

17 18

Table 9: We also assessed the sensitivity of results with re- formance than net income. Further, we fail to find Acknowledgements
Over time changes in the ability of earnings gard to econometric issues. For that purposes, we significant incremental predictive power of aggre- The paper has benefited from comments from Joa-
to explain future firm operating calculated Ramsey’s (1969) Regression Specification gated or individual components of OCI for subse- chim Gassen and participants of the 1st Business
performance Error Test (RESET) which resulted in no rejection quent period’s firm operating performance. The Research Conference ‘Finance, Accounting & Taxa-
with regard to the linear relationship of our models. actuarial gains and losses on defined benefit pension tion’ in Frankfurt (Oder). The comments from two
b
Dependent Raw [Rank] Regression Results Next, we calculated Jarque-Bera (1987) tests for all obligations even seem to merely add noise for the
variable:
a c c anonymous referees were especially helpful. Moreo-
β0 ρ-value β1 ρ-value regression models to see whether the errors are prediction of subsequent period’s net income and of ver, we wish to thank the “Gesellschaft für kapital-
M1 - σε
2
t 0.0000 1.00 0.0014 0.00 distributed normally. These tests showed the kurto- subsequent period’s comprehensive income, respec- marktorientierte Rechnungslegung e. V.” (GKR) for
[0.6667 ] [0.54] [0.87 88] [0.00] sis of the errors to be higher than “3” and/or the tively. A possible reason is found to be that for the providing general support. All remaining errors are
M2 - σε
2
0.0000 0.98 0.0014 0.00
skewness of the errors to be different from “0” for actuarial gains and losses on defined benefit pension the authors’ alone.
t
[1 .2000] [0.41 ] [0.7 81 8] [0.01 ] some models. However, we find these results not to obligations mortality and salary progression are
2 cause main econometric problems due to central estimated, while future plan amendments and future References
M3 - σε t -0.0005 0.65 0.0007 0.00
[0.8000] [0.50] [0.8545] [0.00] limit theorem (CLT) which states that the errors will years of service are not anticipated regardless of the
Basu, Sudipta, The Conservatism Principle and the Asymmetric
2
converge to normality when more random observa- probability of occurrence. In contrast to these find- Timeliness of Earnings, in: Journal of Accounting and Economics,
M4 - σε -0.0004 0.72 0.0007 0.01
t
tions are included. Thus, the errors of our regression ings, we are able to confirm the incremental predic- Vol. 24, Issue 1, 1997, pp. 3-37.
[1 .2667 ] [0.39] [0.7 697 ] [0.01 ]
models are regarded to be approximately normally tive power of OCI components beyond one period. Beaver, William H., Commentary on Problems and Paradoxes in
2
M5 - σε t -0.0003 0.76 0.0007 0.00 distributed (Rice 1995). In summary, the incremental predictive power of the Financial Reporting of Future Events, in: Accounting Hori-
[0.6000] [0.56] [0.8909] [0.00]
Furthermore, we have re-run all regression models OCI components seems to be dependent on the spe- zons, December 1991, pp. 122-134.
2
M6 - σε t -0.0003 0.77 0.0007 0.00 by the application of robust standard errors cific prediction period under examination. Finally, Bernard, Victor L., The Feltham-Ohlson Framework: Implications
[1 .0667 ] [0.44] [0.8061 ] [0.01 ] for Empiricists, in: Contemporary Accounting Research, Vol. 11,
(White [1980]) in order to control for heteroscedas- our analyses show that the predictive power of net
No. 2, Spring 1995, pp. 733-747.
Notes: The panel shows the results of the estimation of models
ticity within our sample. Again, this proceeding re- income and comprehensive for future firm operating
sulted in similar qualitative results. performance seems to have deteriorated as a conse- Biddle, Gary C./ Choi, Jong-Hag, Is Comprehensive Income Use-
that test the association of Gu’s (2007) residual dispersion meas-
ful?, in: Journal of Contemporary Accounting & Economics, Vol. 2,
ure resulting from M1 through to M6 with time. The sample used quence of the IASB’s recent initiatives and actions. No. 1, June 2006, pp. 1-32.
for the cross-sectional regressions consists of all IFRS companies Structural breaks of the over time patterns corres-
belonging to the German major stock exchange index ‘HDAX’ as of
7. Conclusions and limitations Braumoeller, Bear F., Hypothesis Testing and Multiplicative
August 16, 2009. The current study examines the predictive power of pond to the shifts from the IAS period to the IFRS Interaction Terms, in: International Organization, Vol. 58, Issue 4,
aThe residual dispersion measure (σε2) is defined as: σε2 = Σ(εi2/[n- comprehensive income and its individual compo- voluntary period in 2002 and from the IFRS volun- 2004, pp. 807-820.
k]), where εi is the error term of the regression model M1, M2, M3, nents within the homogenous institutional setting of tary period to the IFRS mandatory period in 2004 Breusch, T. S., Testing for Autocorrelation in Dynamic Linear
M4, M5, or M6, respectively, for each observation, n is the number (Paananen and Lin [2009]). Thus, the worsening of Models, in: Australian Economic Papers, Vol. 17, 1978, pp. 334-
of observations of the year and k is the number of independent
German IFRS firms. Specifically, this study wants to
predictive power seems to be driven by main 355.
variables including the intercept term. add empirical evidence to the usefulness of (other)
bWe report coefficient estimates for raw [rank] regressions of: comprehensive income disclosures in comparison to changes of the standards. Choi, Jong-Hag / Das, S. / Zang, Yoonseok, Comprehensive In-
come, Future Earnings and Market Mispricing, Working Paper,
Model M28: σε2t = β0+ β1*t + γt and net income as the hitherto existing bottom-line of The results of this study are subject to limitations.
March 2007.
Model M29: Rank_σε2t = β0+ β1*t + γt Firstly, the study is conducted for German compa-
where σε2t is the residual dispersion measure of the regressions for
the income statement. Such an examination is of Choi, Jong-Hag / Zang, Yoonseok, Implications of Comprehensive
period t, Rank_σε2t is the rank of the residual dispersion measure relevance as the IASB is currently conducting a joint nies only, and therefore, may not be representative Income Disclosure for Future Earning and Analysts’ Forecasts, in:
for the whole time pattern (e. g. “1” for the smallest measure and regulation project together with the FASB in order to for other countries. Secondly, the sample of IFRS Seoul Journal of Business, Vol. 12, No. 2, December 2006, pp. 77-
“10” for the largest measure), and t is the sample year (t = 1…10 enhance the presentation of financial statements. companies belonging to the German major stock 109.
corresponds to years 1998…2007). exchange index ‘HDAX’ as of August 16, 2009 is
cρ-value indicates the one-tailed significance level of the t-statistic
One of the main proposals of this project is to Chow, G. C., Tests of Equality Between Sets of Coefficients in Two

on the relevant variable. present a single statement of comprehensive income relatively small which might influence the infe- Linear Regressions, in: Econometrica, Vol 28, 1960, pp. 591-605.

instead of allowing to report by means of the tradi- rences. Thirdly, our analyses should be taken with Dechow, Patricia M. / Schrand, Catherine M., Earnings Quality,
care as ‘real-world’ investors might consider addi- Charlotteville 2004.
6.6 Sensitivity analyses tional income statement and a reconciliation of net
To examine the sensitivity of the results, several income to comprehensive income within the notes. tional sources beyond accounting information to Dhaliwal, Dan / Subramanyan, K. R. / Trezevant, Robert, Is com-
predict future firm operating performance. Finally, prehensive income superior to net income as a measure of firm
tests were conducted. The tests were performed in We aim to analyze whether such a change in the performance?, in: Journal of Accounting and Economics, Vol. 26,
order to increase the confidence in our main empiri- presentation of the income numbers has the ability we are not able to draw any conclusions whether it 1999, no. 26, pp. 43-67.
cal findings. Firstly, all regression models presented to improve investor’s predictions (based on account- matters if the OCI components are included in com-
D’Mello, Ranjan / Farhat, Joseph, A comparative analysis of
above were run without the use of the dummy varia- ing information) of future firm operating perfor- prehensive income or disclosed elsewhere in the proxies for an optimal leverage ratio, in: Review of Financial
ble and the multiplicative interaction term which mance as represented by future operating cash financial statements (e. g. as a direct adjustment to Economics, Vol. 17, 2008, pp. 213-227.

controls for the timelier recognition of losses in flows, future net income and future comprehensive equity). All these aspects call for future research on Durbin, J. / Watson, G. S., Testing for Serial Correlation in Least
comparison to gains. Secondly, we excluded all income. this field of study. Squares Regression I, in: Biometrika, Vol. 37, 1950, pp. 409-428.

moderating variables from our analyses. Though the This study’s major findings are as follows. We find Durbin, J. / Watson, G. S., Testing for Serial Correlation in Least
Squares Regression II, in: Biometrika, Vol. 38, 1951, pp. 159-178.
adjusted models have less explanatory power, the no evidence that comprehensive income has a supe-
rior predictive power for future firm operating per- Ernstberger, Jürgen, The value relevance of comprehensive in-
results do not change qualitatively.
come under IFRS and US GAAP: empirical evidence from Germa-

19 20
ny, in: International Journal of Accounting, Auditing and Perfor- McClain, Guy / McLelland, Andrew J., Shaking up financial
mance Evaluation, Vol. 5, No. 1, 2008, pp. 1-29. statement presentation, in: Journal of Accountancy, Vol. 206,
No. 5, November 2008, pp. 56-64.
Fama, Eugene F. / French, Kenneth R., Forecasting Profitability
and Earnings, in: Journal of Business, Vol. 73, No. 2, 2000, Miller, Merton H. / Modigliani, Franco, Dividend policy, growth,
pp. 161-175. and the valuation of shares, in: The Journal of Business, Vol. 34,
No. 2, October 1961, pp. 411-433.
Feltham, Gerald A. / Ohlson, James A., Valuation and clean sur-
plus accounting for operating and financial activities, in: Contem- Ohlson, James A., Earnings, book values, and dividends in equity
porary Accounting Research, Vol. 11, No. 2, Spring 1995, pp. 689- valuation, in: Contemporary Accounting Research, Vol. 11, No. 2,
731. Spring 1995, pp. 661-687.
Field, Andy, Discovering Statistics Using SPSS, 2. ed., Lon- Ohlson, James A., On transitory earnings, in: Review of Account-
don 2005. ing Studies, Vol. 4, No. 3-4, 1999, pp. 145-162.
Francis, Jennifer/LaFond, Ryan/Olsson, Per M./Schipper, Kathe- Paananen, Mari / Lin, Henghsiu, The Development of Accounting
rine, Costs of Equity and Earnings Attributes, in: Accounting Quality of IAS and IFRS over Time: The Case of Germany, in:
Review, Vol. 79, Issue 4, 2004, pp. 967-1010. Journal of International Accounting Research, Vol. 8, No. 1,
pp. 31-55.
Francis, J. / Schipper, K., Have Financial Statements Lost Their
Relevance?, in: Journal of Accounting Research, Vol. 37, No. 2, Ramsey, J. B., Tests for Specification Errors in Classical Least
Autumn 1999, pp. 319-352. Squares Regression Analysis, in: Journal of Royal Statistical Socie-
ty, B31, 1969, pp. 350-371.
Godfrey, L. G., Testing Against General Autoregressive and Mov-
ing Average Error Models when the Regressors Include Lagged Rice, John A., Mathematical Statistics and Data Analysis, 2. ed.,
Dependent Variables, in: Econometrica, Vol. 46, No. 6, Novem- Belmont (California) 1995.
ber 1978, pp. 1293-1302.
Thinggaard, Frank / Wagenhofer, Alfred / Evans, Lisa / Gebhardt,
Goncharov, Igor / Hodgson, Allan C., Comprehensive income in Günther / Hoogendorn, Martin / Marton, Jan / di Pietro, Rober-
Europe: Valuation, prediction and conservatism issues, in: An- to / Mora, Araceli / Peasnell, Ken, Performance Reporting – The
nales Universitatis Apulensis Series Oeconomica, Faculty of IASB’s proposed formats of financial statements in the exposure
Sciences, "1 Decembrie 1918" University, Alba Iulia, Vol. 1, Is- draft of IAS 1, in: Accounting in Europe, Vol. 3, 2006, pp. 35-63.
sue 10, October 2008.
Vuong, Quang H., Likelihood ratio tests for model selection and
Greene, William H., Econometric Analysis, 5. ed., New Jer- non-nested hypotheses, in: Econometrica, Vol. 57, No. 2, March
sey 2003. 1989, pp. 307-333.
Gu, Zhaoyang, Across-sample Incomparability of R2s and Addi- Wang, Yue, Essays on the Relevance and Use of Dirty Surplus
tional Evidence of Value Relevance Changes Over Time, in: Jour- Accounting Flows in Europe, Tilburg 2006.
nal of Business Finance & Accounting, Vol. 34, No. 7 & 8, Septem-
White, Halbert, A heteroskedasticity-consistent covariance matrix
ber / October 2007, pp. 1073-1098. estimator and a direct test for heteroskedasticity, in: Econometri-
Hirst, D. Eric / Hopkins, Patrick E., Comprehensive income re- ca, Vol. 48, No. 4, 1980, pp. 817-838.
porting and analysts’ valuation judgments, in: Journal of Account-
ing Research, Vol. 36, Supplement 1998, pp. 47-75.
IASB (ed.), Discussion Paper “Preliminary Views on Financial
Statement Presentation“, October 2008.
IASB (ed.), Reporting comprehensive income, in: IASB update,
July 2003, London 2003, p. 8.
Jarque, C. M. / Bera, A. K., A Test for Normality of Observations
and Regression Residuals, in: International Statistical Review,
Vol. 55, 1987, pp. 163-172.
Kanagaretnam, Kiridaran / Mathieu, Robert / Shehata, Mohamed,
Usefulness of comprehensive income reporting in Canada, in:
Journal of Accounting and Public Policy, Vol. 28, Issue 4, July-
August 2009, pp. 349-365.
Kennedy, Peter, A Guide to Econometrics, 6. ed., Cambridge,
Massachusetts 2008.
Libby, Robert, Accounting and information processing: Theory
and applications, 1st edition, New Jersey, 1981.
Maines, Laureen / McDaniel, Linda S., Effects of Comprehensive
Income Characteristics on Nonprofessional Investors’ Judge-
ments: The Role of Financial-Statement Presentation Format, in:
The Accounting Review, Vol. 75, No. 2, April 2000, pp. 179-207.

21

You might also like