Description
International Accounting Standard 39 (IAS 39) and the Statement of Financial Accounting
Standard 159 (SFAS 159) both require a firm to include adjustments to the fair values of
its liabilities resulting from changes in its own credit risk in net income. Previous research
confirms that including these gains and losses in net income can confuse financial statement
users. The International Accounting Standards Board (IASB) therefore issued a revised
standard for financial instruments, International Financial Reporting Standard (IFRS) 9. It
requires that credit risk effects be presented in other comprehensive income (OCI) instead
of net income. We use an experiment to investigate whether this difference in presentation
affects knowledgeable nonprofessional investors and whether the presentation format
effect depends on firm profitability.
Fair value accounting for liabilities: Presentation format of
credit risk changes and individual information processing
Maik Lachmann
a,?
, Ulrike Stefani
b
, Arnt Wöhrmann
c
a
TU Dortmund University, Department of Accounting & Management Control, Otto-Hahn-Str. 6a, 44227 Dortmund, Germany
b
University of Konstanz, Department of Economics, Box 142, 78457 Konstanz, Germany
c
University of Münster, School of Business and Economics, Universitätsstr. 14-16, 48143 Münster, Germany
a b s t r a c t
International Accounting Standard 39 (IAS 39) and the Statement of Financial Accounting
Standard 159 (SFAS 159) both require a ?rm to include adjustments to the fair values of
its liabilities resulting from changes in its own credit risk in net income. Previous research
con?rms that including these gains and losses in net income can confuse ?nancial state-
ment users. The International Accounting Standards Board (IASB) therefore issued a revised
standard for ?nancial instruments, International Financial Reporting Standard (IFRS) 9. It
requires that credit risk effects be presented in other comprehensive income (OCI) instead
of net income. We use an experiment to investigate whether this difference in presentation
affects knowledgeable nonprofessional investors and whether the presentation format
effect depends on ?rm pro?tability. We ?nd that participants are more likely to acquire
the information on changes in credit risk if that information is included in OCI. The per-
ceived importance of credit risk information for the evaluation of ?rm performance is only
slightly lower under the OCI presentation format, and the risk of misinterpreting a credit
risk gain is unaffected by the presentation format. However, the evaluation of overall ?rm
performance is less biased if fair value gains are included in OCI. Moreover, information
acquisition and the interaction of weighting and credit risk evaluation mediate the effect
of presentation format on ?rm performance evaluation. Furthermore, we identify ?rm
pro?tability as an important factor that in?uences the processing of information on credit
risk changes.
Ó 2014 Elsevier Ltd. All rights reserved.
Introduction
The accounting treatment of changes in a ?rm’s own
credit risk is at the core of a heated debate among academ-
ics, practitioners, and standard setters (e.g., Barth, Hodder,
& Stubben, 2008; FASB, 2006; Gaynor, McDaniel, & Yohn,
2011; IASB, 2009a; Lipe, 2002). Both the Statement of
Financial Accounting Standard No. 159 (SFAS 159) and
International Accounting Standard 39 (IAS 39) allow the
measurement of liabilities at fair value. Because an increase
in a ?rm’s credit risk increases the risk premium charged
on its debt, the liabilities’ net present value decreases and
results in a fair value gain. Similarly, a decrease in credit risk
leads to a fair value loss. According to both SFAS 159 and
IAS 39, credit risk gains and losses are to be recognized
in net income, resulting in a counterintuitive effect: a ?rm’s
worsening ?nancial condition can improve its net income
(European Central Bank, 2001). Because of this concern,
the International Accounting Standards Board (IASB)
issued International Financial Reporting Standard 9 (IFRShttp://dx.doi.org/10.1016/j.aos.2014.08.001
0361-3682/Ó 2014 Elsevier Ltd. All rights reserved.
?
Corresponding author. Tel.: +49 231 755 4373; fax: +49 231 755
3141.
E-mail addresses: [email protected] (M. Lachmann),
[email protected] (U. Stefani), Arnt.Woehrmann@wiwi.
uni-muenster.de (A. Wöhrmann).
Accounting, Organizations and Society 41 (2015) 21–38
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
9) to ‘‘respond to consistent feedback . . . that the effects of
changes in a liability’s credit risk ought not to affect pro?t
or loss’’ (IFRS 9.IN7). IFRS 9 stipulates that changes in the
fair value of liabilities attributable to changes in a ?rm’s
own credit risk must be presented as components of other
comprehensive income (OCI) and that the remaining
amount of the fair value change must be included in net
income (IFRS 9.5.7.7(a)). The assumption made by the IASB
is that the new presentation format helps investors better
interpret the ?rm’s ?nancial performance (IASB, 2010). The
Financial Accounting Standards Board (FASB), in contrast,
came to the opposite conclusion from the IASB. It decided
that there is no justi?cation for constraining the recogni-
tion of the effects resulting from changes in creditworthi-
ness as part of the net income ?gure (FASB, 2007, par.
A25; IASB, 2010). Evidence on whether the new presenta-
tion format implemented by the IASB improves investor
interpretation of a ?rm’s ?nancial performance does not
exist. We investigate this issue in our study.
In addition, the effect of the new presentation format (if
any) likely depends on a ?rm’s operating performance. If a
downgraded ?rm reports a loss, investors are likely sensi-
tized to evaluate the credit risk change more carefully, and
are less likely to interpret a credit risk gain as a positive
signal. If, however, a downgraded ?rm reports a pro?t,
then this cue is missing, and ?rm performance mispercep-
tion becomes more likely. Both pro?t and loss cases are
economically relevant—Jiang (2008) documents that 67%
of the downgraded ?rms report a pro?t, while 33% report
a loss. If the economic and ?nancial condition of a ?rm
indeed differentially affect investors’ perceptions of credit
risk changes, then it is important to examine information
processing in pro?t and loss ?rms separately.
In our study, we experimentally investigate whether
different presentation formats of an increase in a ?rm’s
own credit risk affect knowledgeable nonprofessional
investors’ information processing and whether the effect
of the presentation format depends on ?rm pro?tability.
We apply the framework proposed by Maines and
McDaniel (2000), which separates information processing
into three phases: acquisition (i.e., acquiring, storing, and
recalling information), evaluation (i.e., interpreting an
acquired piece of information), and weighting (i.e., placing
a relative weight on a piece of information). We analyze
whether and how the presentation of credit risk effects
in the OCI statement, instead of in the net income state-
ment, affects these phases of information processing in
pro?t and in loss ?rms.
We implement a 2 Â 2 full factorial between-subjects
design and manipulate the presentation format of credit
risk gains (net income statement versus OCI statement)
and ?rm pro?tability (pro?t versus loss reported). The pre-
sentation format variable re?ects the different approaches
taken by SFAS 159/IAS 39 (net income) and IFRS 9 (OCI). As
participants, we used 93 auditors who served as a proxy for
knowledgeable nonprofessional investors. We provided
them with a ?ctitious ?rm’s ?nancial statements and
recorded their judgments according to the phases of the
Maines and McDaniel (2000) framework.
We ?nd that knowledgeable nonprofessional investors
are more likely to acquire the information on a credit risk
increase if credit risk gains are included in OCI, as required
by IFRS 9. The perceived importance of credit risk informa-
tion for the evaluation of ?rm performance is only slightly
lower when presented in the OCI statement. Most impor-
tantly, we ?nd that the exclusion of credit risk gains and
losses from net income, as required by IFRS 9, prevents
misinterpretations of ?rm performance. However, in the
case of a credit risk increase, the presentation of credit risk
gains in the OCI statement inevitably leads to a lower net
income (and a higher OCI) than under the presentation in
the net income statement required by SFAS 159 / IAS 39.
To rule out the alternative explanation that the lower net
income number in the OCI treatment (instead of the pre-
sentation format) is the primary reason for the improve-
ment in the evaluation of ?rm performance, we conduct
a follow-up experiment to disentangle the presentation
format effects and the effects resulting from differences
in the net income ?gures. We use a structural equations
model to con?rm the importance of the presentation for-
mat by showing that ?rm performance misinterpretation
is the result of differences in acquisition, evaluation, and
weighting of the credit risk information between the
presentation formats. Furthermore, we ?nd that the
differences between the reporting formats in the phases
of acquisition, weighting, and ?rm performance evaluation
are reduced when a loss is reported.
This study contributes to a stream of research dealing
with the relevance of presentation format for the process-
ing of fair value information (e.g. Ahmed, Kilic, & Lobo,
2006; Hirst & Hopkins, 1998). Because liability fair values
depend on credit risk, a change in credit risk can lead to
counterintuitive income effects. Therefore, the ?ndings on
asset fair values do not generalize to the case of liability fair
values (Koonce, Nelson, & Shakespeare, 2011). Since IFRS 9
has not yet been endorsed, empirical studies that use ?eld
data are not yet available. Gaynor et al. (2011) conducted
the only experimental study that provides evidence on the
perception of credit risk effects, and ?nd that more than
70% of the CPAs in their study misinterpret the income
effects of credit risk. Such misinterpretations can provoke
a biased perception of a ?rm’s performance that can induce
inef?cient investment decisions, especially by nonprofes-
sional investors (Chasteen & Ransom, 2007). However, these
authors do not address two important issues.
First, Gaynor et al. (2011, p. 133) ?nd that improved dis-
closures reduce misinterpretation only partially. They
therefore encourage standard setters to ‘‘consider exclud-
ing credit risk gains and losses fromthe income statement.’’
Yet, evidence on the effectiveness of such an approach is
absent. The IASB now requires credit risk gains and losses
to be included in OCI. Our study may therefore aid the Euro-
pean Commission in deciding whether to endorse IFRS 9
and it may inform other standard setters, such as the FASB,
about whether to follow this approach. We also assess the
validity of the IASB’s argument that the new presentation
format should prevent credit risk misinterpretation from
turning into ?rm performance misinterpretation.
Second, Gaynor et al. (2011) focus on a case where the
credit risk effect appears in the current period, while the
underlying economic reason may affect income in a later,
unreported period. In their experiment, ?rm pro?tability
22 M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38
re?ects only the credit risk effect, not the underlying
economic cause. Yet, prior studies showthat a ?rm’s perfor-
mance and its credit risk are intertwined (Ashbaugh-Skaife,
Collins, & LaFond, 2006; Barth, Hodder, et al., 2008; Bhojraj
& Sengupta, 2003). An increase in a ?rm’s credit risk may,
therefore, co-occur with a decrease in its pro?tability.
Because investors do not always treat pro?ts and losses
symmetrically (Bhojraj & Sengupta, 2003), Gaynor et al.’s
(2011) ?ndings for ?rms that report a pro?t might not
generalize to losses, particularly if pro?ts and losses have
different effects onthe perceptionof credit risk information.
Our study considers both pro?t and loss ?rms and thus sets
a boundary condition to Gaynor et al.’s (2011) results.
The rest of the paper proceeds as follows. In the next
section, we describe the background on the debate relating
to the presentation format of fair value gains and losses
due to credit risk changes. We also develop our hypothe-
ses. In the third section, we detail our experimental
procedure and, in the fourth, provide our results. In the
?nal section, we conclude and address the limitations of
our study.
Background and hypotheses development
The debate on reporting credit risk changes
In June 2009, the IASB staff prepared a paper summariz-
ing the pros and cons of including own credit risk in liabil-
ity measurement (IASB, 2009a). The IASB stated the
counterintuitive effect that a decrease in a ?rm’s credit-
worthiness increases net income as the main objection to
the fair value measurement of liabilities according to IAS
39. Comment letters largely con?rmed this view (IASB,
2009b). Respondents to the IASB discussion paper also
raised the concern that including credit risk effects in net
income increases its volatility (IASB, 2010). To mitigate
these problems, the IASB proposed IFRS 9, requiring the
credit risk effect to be included directly in OCI and the
remaining net change in the fair value to be reported in
pro?t or loss (IFRS 9.BC5.45).
Because the income effect of a ?rm’s credit risk changes
has been found to be counterintuitive (Gaynor et al., 2011;
Lipe, 2002), the discussion has focused mainly on whether
knowledgeable ?nancial statement users correctly evalu-
ate a credit risk gain as a negative signal and a loss as a
positive signal. Given the IASB’s proposal, the question
arises as to whether and how the new presentation format
would affect ?nancial statement users’ perceptions of ?rm
performance. We thus investigate the effects of reporting
credit risk changes in the OCI statement. Because previous
research ?nds similar results regarding the effects of credit
risk downgrades and upgrades in the absence of additional
disclosures (Gaynor et al., 2011), we limit our analysis to
the case of downgrades, that is, increases in credit risk.
A general framework for analyzing information processing
Based on the work of Hogarth (1980), Maines and
McDaniel (2000) provide a framework for modeling infor-
mation processing where they disaggregate information
processing into three phases: acquisition, evaluation, and
weighting. During acquisition, a person reads and stores a
piece of information in memory. Only after acquisition
can the reader incorporate the information into an overall
judgment. In addition, the reader must correctly interpret
the information (evaluation). During weighting, the last
step, the reader considers the importance of a particular
piece of information for the pertinent decision.
This framework has been adopted not only to the
reporting of credit risk (Lachmann, Wöhrmann, &
Wömpener, 2011) but also to various accounting contexts
(Elliott, Hodge, Kennedy, & Pronk, 2007; Hirst, Hopkins, &
Wahlen, 2004; Hodge, Kennedy, & Maines, 2004; Nelson
& Tayler, 2007). It also has been applied to various types
of decision makers, including nonprofessional investors
(Maines & McDaniel, 2000) and professional lenders (e.g.,
Hales, Venkataraman, & Wilks, 2011).
Characteristics of the presentation formats expected to
in?uence individual information processing
Credit risk gains are reported as components of net
income according to SFAS 159/IAS 39 and as part of OCI
under IFRS 9. We compare the presentation formats by
using two criteria: (1) the degree of isolation of credit risk
gains and (2) the nature of income of credit risk gains.
1
Building on the similarities and differences between the pre-
sentation formats concerning these criteria, we then develop
our hypotheses for the effects of presentation format in the
phases of information processing.
The ?rst dimension that differs across presentation for-
mats is the degree of isolation of credit risk gains. With an
increased amount of irrelevant information, i.e., with a
lower degree of isolation, it becomes harder to extract
the relevant information (Maines & McDaniel, 2000). Items
presented in the net income statement usually outnumber
those in the statement of comprehensive income.
2
Hence,
the degree of isolation for a single item is lower when
presented in the net income statement and higher when
presented in the OCI statement.
The second criterion for comparison is the nature of
income. Financial statement users will perceive credit risk
gains presented in the net income or the OCI statement dif-
ferently if they regard net income and OCI as elements that
address different kinds of information needs. Thus, equiva-
lent information might be seen differently if users with
limited attention focus on information that is part of a con-
struct considered more important for the question at hand
(Hirshleifer & Teoh, 2003). However, evidence on different
perceptions of the relative importance of net income
versus OCI is scarce. Studies that compare the two income
1
Research on information processing shows that labels (i.e., the deno-
tation of fair value changes as a ‘‘gain’’ or a ‘‘loss’’) can also affect
information processing. However, the presentation formats we analyze do
not differ with respect to labeling.
2
Our analysis of the 2010 annual reports of the Dow Jones Eurostoxx 50
companies con?rms this reasoning: On average, the net income statement
consists of 13.7 items, whereas the OCI statement comprises only 5.7 items
(excluding subtotals). Moreover, the net income statement, on average,
contains 4.7 (sub)totals, whereas the OCI statement has only 1.5
(sub)totals.
M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38 23
concepts usually dwell on the value relevance of special
items included in net income versus OCI gains and losses.
Jones and Smith (2011), for example, show that, although
OCI gains and losses are value relevant, their relevance is
signi?cantly lower than that of special items gains and
losses presented in net income. We therefore conclude that
the relevance of OCI for evaluating ?rm performance is
lower than that of net income.
Importance of net income as contextual information
Firm performance and changes in own credit risk are
closely related (Jiang, 2008). Further, ?rm pro?tability,
which is usually measured by net income (Hodder,
Koonce, & McAnally, 2001; Hopkins, 1996), is an important
contextual dimension that may be relevant for interpreting
credit risk changes. Thus, if ?nancial statement users are
particularly sensitive to changes in creditworthiness due
to changes in net income, differences in the two criteria
between the presentation formats can have a different
impact on information processing.
In this study, we compare information processing
between the net income and the OCI presentation formats
for situations in which the ?rm reports a pro?t and for sit-
uations in which a loss is realized. Prior experimental
research has sidestepped the moderating effects of changes
in a ?rm’s economic and ?nancial conditions by assuming
that the change in credit risk materializes before the
change in the ?rm’s economic condition does. In Gaynor
et al.’s (2011) experiment, for example, the fair value of lia-
bilities is affected in the current period, but the reason for
the credit risk change (i.e., the signing or the loss of a long-
term contract) in?uences income only in the following
(unreported) period. In other words, current-period pro?t-
ability does not re?ect the economic cause of the credit
risk change. This situation precludes a diagnosis of the rea-
sons that led to the credit risk increase, which is essential
for causal reasoning (Koonce, Seybert, & Smith, 2011).
Whether an increase in credit risk materializes before a
?rm’s economic condition deteriorates is a questionable
assumption. It is equally likely that rating downgrades
coincide with (or promptly follow) the worsening of a
?rm’s economic condition. Timing matters here. If inves-
tors are aware of the deterioration of a ?rm’s economic
condition, they may read the credit risk information more
diligently. Investors certainly assume an extremely nega-
tive change in the economic condition when net income
turns from positive to negative (e.g., Barth, Landsman, &
Lang, 2008; Burgstahler & Dichev, 1997; Leuz, Nanda, &
Wysocki, 2003).
Hypotheses development
Acquisition of credit risk information
The ?rst step in the information-processing framework
presented above is acquisition, which not only requires
reading but also storing and recalling information (Elliott
et al., 2007). Once the information is read, it must then
be represented ‘‘internally in a way that enables its use’’
(Hirshleifer & Teoh, 2003, p. 342). Internal representation
is affected by cognitive costs (Hirshleifer & Teoh, 2003)
and an item’s salience (Frederickson & Miller, 2004).
Hodge et al. (2004) ?nd that presentation format affects
acquisition. Because the degree of isolation is higher in
the OCI presentation format and a higher degree of isola-
tion leads to lower cognitive costs (Maines & McDaniel,
2000), individuals are more likely to acquire information
on credit risk gains if it is presented in the OCI rather than
in the net income statement. Consequently, we posit the
following hypothesis:
H1a. Individuals are more likely to acquire information on
credit risk changes when gains from credit risk increases
are presented in the OCI statement rather than in the net
income statement.
The earnings management literature shows that man-
agers try to avoid reporting small losses (Burgstahler &
Chuk, 2013; Burgstahler & Dichev, 1997; Jiang, 2008)
because losses are interpreted as ‘‘red ?ags’’. When ?nan-
cial statement users interpret a negative net income as a
warning, they might be more likely to read the ?nancial
information more diligently and to dedicate more atten-
tion to the components of net income to identify the reasons
for the loss. This will have a positive effect on the acquisi-
tion of credit risk changes when credit risk gains are
reported in the net income statement (where the loss is
reported) whereas acquisition is unaffected when credit
risk gains are reported in the OCI statement. In other
words, when a loss is reported in the income statement,
the likelihood of acquiring information on credit risk
changes increases under the net income presentation for-
mat but not under the OCI presentation format. Hence,
the relative advantage of OCI presentation, which stems
from its higher degree of isolation of information on credit
risk gains (H1a), should be reduced when a loss is reported.
However, we do not expect that the differences between
the presentation formats are completely eliminated when
a loss is reported because the net income statement
contains more items than the OCI statement, i.e., acquisi-
tion under the net income presentation format is still more
costly than under the OCI presentation format. We
therefore posit the following hypothesis:
H1b. Differences in the likelihood of acquiring information
on credit risk changes presented in the OCI statement
versus in the net income statement are reduced when a
loss is reported.
Panel A of Fig. 1 graphically depicts our predictions for
acquiring information on credit risk changes (H1a and
H1b).
Evaluation and weighting of information about credit risk
changes
Once ?nancial statement users acquire information on
credit risk gains, they evaluate and weigh it. Gaynor et al.
(2011) argue that users typically interpret gains as positive
and losses as negative signals (Hodder et al., 2001). Yet,
this intuition is misleading in the case of fair value adjust-
ments resulting from changes in own credit risk. However,
a credit risk increase leads to a gain under both presenta-
tion formats; we therefore expect no differences in the risk
24 M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38
of misinterpretation (i.e., evaluation) between the
presentation formats.
Presentation format can affect weighting in two
different and opposite ways. On the one hand, H1a states
that—because of the lower degree of isolation in the net
income presentation format—users are less likely to
acquire a credit risk change when credit risk gains are
reported in the net income statement. The degree of isola-
tion also affects the weighting phase: An investor who
does not acquire the credit risk change will place a low
weight on credit risk information when assessing ?rm
performance. Put differently, the lower degree of isolation
in the net income presentation format undermines
weighting (Nisbett, Zukier, & Lemley, 1981).
Conversely, the nature of income criterion suggests that
the user’s perception of the value relevance of credit risk
gains would be higher when these gains were presented
in the net income statement. Net income items are
regarded as better predictors of ?rm performance than
OCI items (Hodder et al., 2001; Hopkins, 1996; Jones &
Smith, 2011). Nonprofessional investors with limited cog-
nitive abilities or limited time will therefore focus more
on information contained in the net income statement
when assessing ?rm performance. Maines and McDaniel
(2000) also argue that a closer link to net income increases
the weight that nonprofessionals give to an income
component. Thus, the nature of income criterion results
in a positive effect on weighting when credit risk gains
appear in the net income statement.
Because the degree of isolation and the nature of
income pull in opposite directions, the overall effect of
the presentation format on weighting is unclear. To derive
a directional hypothesis, we focus only on those users who
have managed to acquire the change in credit risk and are
therefore not negatively affected by the lower degree of
isolation in the net income presentation format. The
weight they place on credit risk information is thus deter-
mined only by the nature of income. We predict that users
who have successfully acquired the change in credit risk
will place more weight on the corresponding information
when it is presented in the net income statement than
when presented in the OCI statement.
H2a. Presenting credit risk gains in the OCI statement,
rather than in the net income statement, decreases the
weight individuals place on previously acquired credit risk
information when assessing overall ?rm performance.
If the ?rm reports a loss, investors pay more attention
to identifying the reasons for the loss and less attention
to analyzing the gains that result from changes in the fair
value of liabilities. This shift in attention reduces the
weight placed on credit risk information. We expect that
the weights placed on credit risk gains are reduced to a lar-
ger extent when credit risk gains are presented in the net
income statement than when presented in the OCI
statement (where we expect no or only a small change in
the weights because the weights under the OCI presenta-
tion format are predicted to be already comparatively
low). Thus, the shift in attention due to the presence of a
loss should reduce – although not eliminate – the
differences in weighting between the presentation
formats. This leads to our next hypothesis.
H2b. Differences in the weighting of previously acquired
credit risk information presented in the OCI statement
versus in the net income statement are reduced when a
loss is reported.
Panel B of Fig. 1 graphically depicts our predictions for
weighting credit risk gains (H2a and H2b).
Overall evaluation of ?rm performance
In releasing IFRS 9, the IASB aimed to reduce net income
volatility and the likelihood of ?nancial statement users
misinterpreting ?rm performance (IASB, 2010). The IASB
did not argue that incorporating credit risk gains in OCI
would reduce credit risk misinterpretation but rather that
it would reduce ?rm performance misinterpretation.
Whereas the total amount of comprehensive income does
not depend on the presentation format, net income is
higher if it contains the credit risk gain (and OCI is lower
Panel A: Predicted pattern for H1
Panel B: Predicted pattern for H2
Panel C: Predicted pattern for H3
Fig. 1. Predictions for H1, H2, and H3.
M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38 25
by the same amount). Hence, all else equal, the presenta-
tion format leads to differences in the values of net income
and OCI.
The correct evaluation of ?rmperformance requires that
the investor mathematically subtract (i.e., neutralize) the
credit risk gain from income to calculate an ‘‘as if’’ income
that is not in?ated by the gain. Without this adjustment,
the investor’s estimate of ?rm performance is biased
upwards; that is, ?rm performance is misinterpreted. We
predict that the acquisition, evaluation, and weighting of
credit risk information determine whether the investor is
likely to neutralize the credit risk gain. We identify two
alternative cases in which an investor does not neutralize a
credit risk gain and misinterprets ?rm performance. As
explained below, we predict that both cases are more likely
whencredit risk gains are presentedinthe net income state-
ment. We do not require both arguments to simultaneously
hold for a speci?c investor, that is, we do not aggregate both
factors. Instead, we predict that each factor can separately
lead to the misinterpretation of ?rm performance.
First, some investors may not have acquired the credit
risk effect on income; these investors cannot neutralize
the gain because they are not aware of the relevant infor-
mation. When evaluating ?rm performance, they will rely
instead on the reported income measure, in?ated by the
fair value gain. H1a predicts that the acquisition of credit
risk changes is more likely when credit risk gains are
presented in the OCI statement. This should enable more
investors to neutralize the credit risk income effect.
Misinterpretation of ?rm performance should thus be less
likely in the OCI presentation format.
Second, some of those investors who have acquired the
credit risk information (and evaluated and weighted the
credit risk gain) still will not neutralize its effect. This
behavior is likely to occur for those investors who miscon-
strue the gain as a positive signal and consider it as highly
relevant for the evaluation of ?rm performance. We do not
make a prediction regarding differences in the risk of mis-
interpreting credit risk gains between the two presentation
formats, but we expect differences in weighting: H2a pre-
dicts that users will place less weight on the acquired
information on credit risk changes if the gains are pre-
sented as an OCI item (Hodder et al., 2001; Hopkins,
1996). If, in contrast, net income is used as a primary
source for the assessment of ?rm performance, then inves-
tors will place more weight on (potentially misinterpreted)
credit risk gains because net income is seen as a key pre-
dictor of the ?rm’s future development. We predict there-
fore that, when credit risk gains are reported in the net
income statement, fewer investors will neutralize the
credit risk income effect, and that investors will be more
likely to misinterpret ?rm performance.
In summary, we offer two mechanisms by which an
investor is less likely to neutralize a credit risk gain when
credit risk gains are presented in the net income than in
the OCI statement. We make our formal prediction below.
H3a. Individuals are less likely to make biased estimates of
overall ?rm performance when credit risk gains are
presented in the OCI statement rather than in the net
income statement.
If net income turns negative, then the change in a ?rm’s
economic condition becomes obvious. Even if users regard
the credit risk gain as a positive signal, the overall effect on
net income is negative. Therefore, we hypothesize that the
difference between the presentation formats regarding
overall ?rm performance evaluation is reduced in the pres-
ence of a loss. We expect that this decrease in the differ-
ence is mainly driven by less biased ?rm performance
evaluations under the net income presentation format.
For the OCI presentation format, in contrast, we expect
no (or only a small) change regarding the likelihood of
biased ?rm performance evaluations. The reasons are that
net income is not in?ated by the credit risk gain in the OCI
presentation format, and that OCI is seen as less predictive
for the ?rm’s future development than net income. Since
investors rely on net income for their assessment of ?rm
performance and net income is higher by the amount of
the fair value gain in the net income presentation format,
we do not expect that the differences in ?rm performance
evaluation between the presentation formats are
completely eliminated when a loss is reported.
H3b. Differences in the likelihood of making biased esti-
mates of overall ?rm performance between the presenta-
tion formats are reduced when a loss is reported.
Panel C of Fig. 1 graphically depicts our predictions for
?rm performance evaluation (H3a and H3b). The lower
part of Fig. 2 shows how the information-processing
framework relates to our hypotheses.
Experimental method
Design
We conducted an experiment using a 2 Â 2 between-
subjects full-factorial design. The ?rst treatment variable
was the presentation format of a fair value gain resulting
from an increase in credit risk. This gain was either
included in the net income statement (NI), as required by
SFAS No. 159 and IAS 39, or in the other comprehensive
income statement (OCI), as required by IFRS 9. Under both
conditions, the credit risk increase led to a gain of
€1,920,000. The notes informed participants that the fair
value gain was attributable entirely to a change in the
?rm’s credit risk. However, the notes did not provide infor-
mation on the direction of the credit risk change.
3
The second treatment variable was whether a negative
pro?tability development (LOSS) occurred simultaneously
with the downgrade or not (PROFIT). Asset impairments
are a common reason for net income to turn negative when
economic conditions deteriorate (Barth, Landsman, et al.,
2008). In the LOSS treatment, an R&D impairment of
€3,900,000 overcompensated for the positive income effect
from the fair value gain. Thus, both net income and
3
Neither GAAP nor IFRS require disclosure of an explanation regarding
the direction of the credit risk change. Thus, preparers usually do not
provide such disclosures. We analyzed the 2011 annual reports of the 10
largest European and 10 largest US banks. We did not ?nd even one
institution that disclosed the direction of a credit risk change when
providing information on the calculation of credit risk gains/losses.
26 M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38
comprehensive income were negative in the LOSS treat-
ment and positive in the PROFIT treatment. All remaining
?nancial items were identical for the LOSS and PROFIT
conditions.
Participants
We used 93 auditors as participants. This allows us to
compare our results to the ?ndings of Gaynor et al.
(2011) who use CPAs. We regard auditors as an appropriate
proxy for knowledgeable nonprofessional investors
because they have a reasonable degree of ?nancial knowl-
edge, as required by the FASB and the IASB, to understand
?nancial statements (Concepts Statement No. 8, IASB
Conceptual Framework).
4
The professional activity of 88% of our participants was
auditing. The remaining 12% worked in the ?eld of audit-
related services. Ninety-?ve percent worked for audit
?rms. Thirty-two percent of these participants were
employed by one of the Big 4, and 68% by non-Big 4 audit
?rms. On average, our participants had 6.8 years of profes-
sional experience and were 34.3 years old.
5
They also had,
on average, 4.9 years of experience in making private invest-
ment decisions. We randomly assigned our participants to
the four treatment conditions NI_PROFIT (n = 22), NI_LOSS
(n = 24), OCI_PROFIT (n = 24), and OCI_LOSS (n = 23). We did
not detect any signi?cant differences between the four
treatment groups with respect to personal characteristics
(all p-values > 0.1).
Procedure and payments
We ran a paper-and-pencil experiment during a profes-
sional training seminar.
6
The experiment took 45 minutes
on average. We gave participants printed instructions con-
taining a brief description of the accounting rules. Partici-
pants under the NI condition learned from these rules that
gains and losses resulting from credit risk changes were pre-
sented in the net income statement, whereas participants
under the OCI condition learned that these gains and losses
were presented in the OCI statement. After reading the
instructions, participants received three sealed envelopes
at different times during the experiment. The envelopes con-
tained materials and question sets that we asked our partic-
ipants to complete consecutively.
The participants received a ?xed fee of €7 and a perfor-
mance-contingent payment between €0 and €10. During
the experiment, we rewarded them with the experimental
currency ‘‘Points’’. At the end of the experiment, we con-
verted Points into euros (100 Points = €1 US$1.3 at the
time of the experiment) and anonymously paid them in
cash. They earned 100 Points for a correct answer and lost
100 Points for a wrong answer.
Because risk preferences can affect perceptions of cred-
itworthiness, we elicited our participants’ risk preferences
Fig. 2. Hypotheses, related questions, and the information processing framework.
4
Though many studies use MBA students as a proxy for nonprofessional
investors (Elliott et al., 2007), some use CPAs, even though the experimen-
tal task is an investment task rather than an auditing task (e.g.,
Frederickson, Hodge, & Pratt, 2006; Gaynor et al., 2011; Ghani, Laswad, &
Tooley, 2011). Gaynor et al. (2011) argue that understanding credit risk
information is complex and that MBA students might not be sophisticated
enough to interpret this information properly.
5
Unlike in the United States, where two years of work experience is
usually required before becoming a CPA, in our case—barring minor
exceptions—graduates need three years and undergraduates need four
years of work experience before being admitted to the exam.
6
The experiment was conducted in Germany. A professional editing
service translated all the excerpts from the instrument that are contained in
this paper into English, and we double-checked the translation. Since the
IFRS are available both in English and in German, we used the of?cial
translation where possible.
M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38 27
using a lottery task (Dohmen et al., 2011; Sprinkle,
Williamson, & Upton, 2008). They had to choose, for 20
cases, between a lottery with equally probable opportuni-
ties to earn 300 Points or zero Points and a safe payment
s (0 6 s 6 190 Points). We followed Dohmen et al. (2011)
and classi?ed participants based on their choices as risk
averse, risk neutral, or risk taking and added the payoff
earned in the lottery task to the participants’ compensation.
We also included three problems from the Cognitive
Re?ection Test (CRT) suggested by Frederick (2005) to eli-
cit a speci?c cognitive ability. All of the questions from the
CRT provoke an intuitive but wrong answer, whereas the
correct answer requires more re?ection. The CRT problems
and our experimental task require similar abilities because
interpreting a credit risk gain as a positive signal is also
intuitive but incorrect. We did not provide monetary
incentives for completing the CRT.
Total compensation varied from €9 to €20, with an
average of €17.59. Participants in the NI_PROFIT/NI_LOSS/
OCI_PROFIT/OCI_LOSS conditions received €16.96/€18.02/
€17.40/€17.94. There was no signi?cant difference in
compensation among the four groups (Kruskal–Wallis test,
v
2
= 1.45, p = 0.69).
Task and instrument
Participants had ?ve minutes to read the instructions
and the accounting rules. They then received the ?nancial
statement of the ?ctitious Company XY with data for the
current and the preceding period (balance sheet, statement
of comprehensive income, cash ?ow statement, and foot-
note disclosures). Panel A of the appendix summarizes
the key ?nancial information. Re?ecting our observations
of business practice, the degree of isolation for credit risk
information was higher under the OCI than under the NI
treatment conditions. In the OCI treatments, the credit risk
gain was one of four entries included in the OCI statement,
while in the NI treatment it was one of eight.
Participants had eight minutes to read the ?nancial
statements before we handed out the ?rst envelope, con-
taining two questions. The ?rst question asked for an esti-
mate of the stock price change in the current period
(relative to the previous year).
7
Speci?cally, participants
had to estimate on a scale from À11 to +11 whether the
stock price had declined considerably (À11), remained
unchanged (0), or increased considerably (+11) in the cur-
rent year relative to the preceding year. If no credit risk gain
had occurred in the current period, then the deterioration of
?rm performance would have been obvious under all four
conditions. Investors who did not neutralize the credit risk
gain from income were misled to provide a biased, that is,
too positive stock price estimate. Participants’ stock price
estimates were used to test H3 (?rm performance evalua-
tion). The second question asked participants to rate the
relevance of ?ve factors (return on equity, cash ?ow, compo-
sition of earnings and expenses, creditworthiness, and capi-
tal structure) when evaluating the stock price of Company
XY by using a scale from 0 (not at all important) to 11 (extre-
mely important). We used the answers to this question to
test H2 (weighting).
We then collected the materials and the ?nancial state-
ments from participants before giving them a second set of
questions. This time, the ?rst question asked whether
Company XY had experienced a change in its creditworthi-
ness thus verifying whether our participants had acquired
the change in creditworthiness (H1) (acquisition). The sec-
ond question asked whether there had been a change in
the fair value of long-term debt. We used this question
as a supplemental test. We then collected these materials
back and handed out the last set of questions.
Participants could again refer to the ?nancial state-
ments when completing this third set of questions. Their
answers to these additional questions were not used for
testing our hypotheses but to better understand their
behavior. First, we asked them to imagine holding a portfo-
lio of shares of ten different companies, including shares of
Company XY. We then instructed them to buy additional
shares if the ?rm’s creditworthiness had increased and to
sell all of the shares if the ?rm’s creditworthiness had
decreased. We asked which option (buy or sell) they would
choose if their decision exclusively depended on the
change in the ?rm’s creditworthiness. Following Gaynor
et al. (2011), we used this question as a manipulation
check to determine whether participants correctly inter-
preted the credit risk change (evaluation of creditworthi-
ness). We asked them to brie?y explain their decision in
writing. As a supplemental test, we requested an estimate
of the magnitude of the change in creditworthiness on a
scale from 0 (no change) to 11 (very strong change).
Fig. 2 presents the relations between the questions, our
hypotheses, and the phases in the information-processing
framework.
Finally, we applied the lottery task and the CRT. At the
end of the experiment, participants answered manipula-
tion check questions and provided demographic data.
Results
Manipulation and other checks
The ?rst manipulation check question asked whether
fair value gains and losses occurring from changes in
own credit risk were included in net income or in OCI. In
both the OCI and NI treatments, 95.7% of our participants
responded correctly. Next, we asked whether there was
an impairment of an R&D project in the current period.
For the PROFIT (LOSS) treatments, 91.3% (100%) of our par-
ticipants answered this question correctly. These results
indicate that both manipulations were successful.
8
7
Since participants had to answer the questions concerning a ?ctitious
company and we wanted to rule out assumptions about future develop-
ments, we asked for an estimate of the stock price of the current period
relative to the previous period. If we had asked for the stock price of a future
period, participants might have made different assumptions concerning the
predictive power of the (transitory) loss under the LOSS conditions.
8
The results presented in the following sections are based on the full
sample. Unless stated otherwise, the results are identical if we exclude
participants who responded incorrectly to the manipulation check
questions.
28 M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38
As stated above, we inferred our participants’ interpre-
tations of the credit risk gain by informing them that they
should sell all of their shares of the ?ctitious ?rm when
credit risk had increased. We take the participants’ deci-
sions not to sell the shares as an indication of a misinter-
pretation of the credit risk gain. To verify that our
participants correctly understood this question, we asked
them the same question for a different ?rm, Company Z,
but told them that the credit risk of this ?rmhad increased.
Some 97% of our participants correctly answered that they
should sell the shares of Company Z.
We elicited risk preferences using the lottery task but
could not classify 17% of our participants because they pro-
vided inconsistent responses across the different lotteries.
Of the remaining, we found that 43% were risk neutral, 43%
were risk averse, and 14% were risk taking. We also mea-
sured their cognitive abilities by using the number of cor-
rect answers to the CRT questions, that is, the CRT score.
Because some of the participants did not answer the sec-
ond and/or third question of the CRT (17% and 25%, respec-
tively), we decided to consider only the ?rst CRT question,
which only one participant did not answer.
9
A total of 40%
of participants answered the ?rst question correctly. The
results from the lottery task and the ?rst question of the
CRT show no statistically signi?cant differences regarding
risk aversion or cognitive abilities between the four treat-
ment groups (p-values > 0.1).
10
Test of H1 (acquisition)
Hypothesis 1a states that the likelihood of acquiring
credit risk information is higher when credit risk gains
are reported in OCI. As presented in Panel A of Table 1,
we ?nd that, on average, 74% of our participants under
the NI conditions and 89% under the OCI conditions cor-
rectly recalled the change in creditworthiness. In line with
our prediction in H1a, we ?nd that the likelihood of acquir-
ing credit risk information is higher when credit risk gains
are reported in OCI (v
2
= 4.38, p = 0.04), as depicted in
Panel B of Table 1. Panel A of Fig. 3 compares our predic-
tions and our results for H1.
When the ?ctitious ?rm reported a pro?t, 59% (88%) in
the NI_PROFIT (OCI_PROFIT) cell acquired the credit risk
change (i.e., the difference across the NI and the OCI treat-
ments in the percentage of participants acquiring this
information is 29%). This difference is signi?cant (chi-
squared-test, v
2
= 4.80, p = 0.03, two-tailed). Yet, when
the ?ctitious ?rm reported a loss, 88% (91%) of the partic-
ipants in the NI_LOSS (OCI_LOSS) cell acquired the credit
risk change; i.e., the difference in acquisition falls to 3%
and becomes insigni?cant (Fisher’s exact test, p = 1.00,
two-tailed). This ?nding is in line with H1b that posits that
differences in the likelihood of acquiring information on
credit risk changes between the presentation formats are
reduced when a loss is reported. However, the ANOVA
interaction term NI Â LOSS in Panel B of Table 1 is insignif-
icant (v
2
= 0.92, p = 0.34). To further analyze the functional
form of the interaction, we test H1b using planned
contrasts, which has two advantages (Buckless &
Ravenscroft, 1990). First, contrast codes increase the statis-
tical power without increasing Type I error rates. Second,
we predict an ordinal interaction. Consistent with our pre-
diction, we use contrast weights of À3 for the NI _PROFIT
condition, À1 for the NI _LOSS condition, and +2 for the
OCI conditions (i.e., OCI_PROFIT and OCI_LOSS) (see Fig. 3,
Table 1
Acquisition (H1).
Pro?tability Presentation format
P
NI OCI
Panel A: Descriptive statistics
Dependent variable: Change in credit risk acquired (binary)
PROFIT 59% (n = 22) 88% (n = 24) 74%
LOSS 88% (n = 24) 91% (n = 23) 89%
P
74% 89% 82%
Source df v
2
p-Value
Panel B: Categorical modeling statistics (logistic regression)—test of H1a (n = 93)
Dependent variable: Change in credit risk acquired (binary)
Intercept 1 9.94
International Accounting Standard 39 (IAS 39) and the Statement of Financial Accounting
Standard 159 (SFAS 159) both require a firm to include adjustments to the fair values of
its liabilities resulting from changes in its own credit risk in net income. Previous research
confirms that including these gains and losses in net income can confuse financial statement
users. The International Accounting Standards Board (IASB) therefore issued a revised
standard for financial instruments, International Financial Reporting Standard (IFRS) 9. It
requires that credit risk effects be presented in other comprehensive income (OCI) instead
of net income. We use an experiment to investigate whether this difference in presentation
affects knowledgeable nonprofessional investors and whether the presentation format
effect depends on firm profitability.
Fair value accounting for liabilities: Presentation format of
credit risk changes and individual information processing
Maik Lachmann
a,?
, Ulrike Stefani
b
, Arnt Wöhrmann
c
a
TU Dortmund University, Department of Accounting & Management Control, Otto-Hahn-Str. 6a, 44227 Dortmund, Germany
b
University of Konstanz, Department of Economics, Box 142, 78457 Konstanz, Germany
c
University of Münster, School of Business and Economics, Universitätsstr. 14-16, 48143 Münster, Germany
a b s t r a c t
International Accounting Standard 39 (IAS 39) and the Statement of Financial Accounting
Standard 159 (SFAS 159) both require a ?rm to include adjustments to the fair values of
its liabilities resulting from changes in its own credit risk in net income. Previous research
con?rms that including these gains and losses in net income can confuse ?nancial state-
ment users. The International Accounting Standards Board (IASB) therefore issued a revised
standard for ?nancial instruments, International Financial Reporting Standard (IFRS) 9. It
requires that credit risk effects be presented in other comprehensive income (OCI) instead
of net income. We use an experiment to investigate whether this difference in presentation
affects knowledgeable nonprofessional investors and whether the presentation format
effect depends on ?rm pro?tability. We ?nd that participants are more likely to acquire
the information on changes in credit risk if that information is included in OCI. The per-
ceived importance of credit risk information for the evaluation of ?rm performance is only
slightly lower under the OCI presentation format, and the risk of misinterpreting a credit
risk gain is unaffected by the presentation format. However, the evaluation of overall ?rm
performance is less biased if fair value gains are included in OCI. Moreover, information
acquisition and the interaction of weighting and credit risk evaluation mediate the effect
of presentation format on ?rm performance evaluation. Furthermore, we identify ?rm
pro?tability as an important factor that in?uences the processing of information on credit
risk changes.
Ó 2014 Elsevier Ltd. All rights reserved.
Introduction
The accounting treatment of changes in a ?rm’s own
credit risk is at the core of a heated debate among academ-
ics, practitioners, and standard setters (e.g., Barth, Hodder,
& Stubben, 2008; FASB, 2006; Gaynor, McDaniel, & Yohn,
2011; IASB, 2009a; Lipe, 2002). Both the Statement of
Financial Accounting Standard No. 159 (SFAS 159) and
International Accounting Standard 39 (IAS 39) allow the
measurement of liabilities at fair value. Because an increase
in a ?rm’s credit risk increases the risk premium charged
on its debt, the liabilities’ net present value decreases and
results in a fair value gain. Similarly, a decrease in credit risk
leads to a fair value loss. According to both SFAS 159 and
IAS 39, credit risk gains and losses are to be recognized
in net income, resulting in a counterintuitive effect: a ?rm’s
worsening ?nancial condition can improve its net income
(European Central Bank, 2001). Because of this concern,
the International Accounting Standards Board (IASB)
issued International Financial Reporting Standard 9 (IFRShttp://dx.doi.org/10.1016/j.aos.2014.08.001
0361-3682/Ó 2014 Elsevier Ltd. All rights reserved.
?
Corresponding author. Tel.: +49 231 755 4373; fax: +49 231 755
3141.
E-mail addresses: [email protected] (M. Lachmann),
[email protected] (U. Stefani), Arnt.Woehrmann@wiwi.
uni-muenster.de (A. Wöhrmann).
Accounting, Organizations and Society 41 (2015) 21–38
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
9) to ‘‘respond to consistent feedback . . . that the effects of
changes in a liability’s credit risk ought not to affect pro?t
or loss’’ (IFRS 9.IN7). IFRS 9 stipulates that changes in the
fair value of liabilities attributable to changes in a ?rm’s
own credit risk must be presented as components of other
comprehensive income (OCI) and that the remaining
amount of the fair value change must be included in net
income (IFRS 9.5.7.7(a)). The assumption made by the IASB
is that the new presentation format helps investors better
interpret the ?rm’s ?nancial performance (IASB, 2010). The
Financial Accounting Standards Board (FASB), in contrast,
came to the opposite conclusion from the IASB. It decided
that there is no justi?cation for constraining the recogni-
tion of the effects resulting from changes in creditworthi-
ness as part of the net income ?gure (FASB, 2007, par.
A25; IASB, 2010). Evidence on whether the new presenta-
tion format implemented by the IASB improves investor
interpretation of a ?rm’s ?nancial performance does not
exist. We investigate this issue in our study.
In addition, the effect of the new presentation format (if
any) likely depends on a ?rm’s operating performance. If a
downgraded ?rm reports a loss, investors are likely sensi-
tized to evaluate the credit risk change more carefully, and
are less likely to interpret a credit risk gain as a positive
signal. If, however, a downgraded ?rm reports a pro?t,
then this cue is missing, and ?rm performance mispercep-
tion becomes more likely. Both pro?t and loss cases are
economically relevant—Jiang (2008) documents that 67%
of the downgraded ?rms report a pro?t, while 33% report
a loss. If the economic and ?nancial condition of a ?rm
indeed differentially affect investors’ perceptions of credit
risk changes, then it is important to examine information
processing in pro?t and loss ?rms separately.
In our study, we experimentally investigate whether
different presentation formats of an increase in a ?rm’s
own credit risk affect knowledgeable nonprofessional
investors’ information processing and whether the effect
of the presentation format depends on ?rm pro?tability.
We apply the framework proposed by Maines and
McDaniel (2000), which separates information processing
into three phases: acquisition (i.e., acquiring, storing, and
recalling information), evaluation (i.e., interpreting an
acquired piece of information), and weighting (i.e., placing
a relative weight on a piece of information). We analyze
whether and how the presentation of credit risk effects
in the OCI statement, instead of in the net income state-
ment, affects these phases of information processing in
pro?t and in loss ?rms.
We implement a 2 Â 2 full factorial between-subjects
design and manipulate the presentation format of credit
risk gains (net income statement versus OCI statement)
and ?rm pro?tability (pro?t versus loss reported). The pre-
sentation format variable re?ects the different approaches
taken by SFAS 159/IAS 39 (net income) and IFRS 9 (OCI). As
participants, we used 93 auditors who served as a proxy for
knowledgeable nonprofessional investors. We provided
them with a ?ctitious ?rm’s ?nancial statements and
recorded their judgments according to the phases of the
Maines and McDaniel (2000) framework.
We ?nd that knowledgeable nonprofessional investors
are more likely to acquire the information on a credit risk
increase if credit risk gains are included in OCI, as required
by IFRS 9. The perceived importance of credit risk informa-
tion for the evaluation of ?rm performance is only slightly
lower when presented in the OCI statement. Most impor-
tantly, we ?nd that the exclusion of credit risk gains and
losses from net income, as required by IFRS 9, prevents
misinterpretations of ?rm performance. However, in the
case of a credit risk increase, the presentation of credit risk
gains in the OCI statement inevitably leads to a lower net
income (and a higher OCI) than under the presentation in
the net income statement required by SFAS 159 / IAS 39.
To rule out the alternative explanation that the lower net
income number in the OCI treatment (instead of the pre-
sentation format) is the primary reason for the improve-
ment in the evaluation of ?rm performance, we conduct
a follow-up experiment to disentangle the presentation
format effects and the effects resulting from differences
in the net income ?gures. We use a structural equations
model to con?rm the importance of the presentation for-
mat by showing that ?rm performance misinterpretation
is the result of differences in acquisition, evaluation, and
weighting of the credit risk information between the
presentation formats. Furthermore, we ?nd that the
differences between the reporting formats in the phases
of acquisition, weighting, and ?rm performance evaluation
are reduced when a loss is reported.
This study contributes to a stream of research dealing
with the relevance of presentation format for the process-
ing of fair value information (e.g. Ahmed, Kilic, & Lobo,
2006; Hirst & Hopkins, 1998). Because liability fair values
depend on credit risk, a change in credit risk can lead to
counterintuitive income effects. Therefore, the ?ndings on
asset fair values do not generalize to the case of liability fair
values (Koonce, Nelson, & Shakespeare, 2011). Since IFRS 9
has not yet been endorsed, empirical studies that use ?eld
data are not yet available. Gaynor et al. (2011) conducted
the only experimental study that provides evidence on the
perception of credit risk effects, and ?nd that more than
70% of the CPAs in their study misinterpret the income
effects of credit risk. Such misinterpretations can provoke
a biased perception of a ?rm’s performance that can induce
inef?cient investment decisions, especially by nonprofes-
sional investors (Chasteen & Ransom, 2007). However, these
authors do not address two important issues.
First, Gaynor et al. (2011, p. 133) ?nd that improved dis-
closures reduce misinterpretation only partially. They
therefore encourage standard setters to ‘‘consider exclud-
ing credit risk gains and losses fromthe income statement.’’
Yet, evidence on the effectiveness of such an approach is
absent. The IASB now requires credit risk gains and losses
to be included in OCI. Our study may therefore aid the Euro-
pean Commission in deciding whether to endorse IFRS 9
and it may inform other standard setters, such as the FASB,
about whether to follow this approach. We also assess the
validity of the IASB’s argument that the new presentation
format should prevent credit risk misinterpretation from
turning into ?rm performance misinterpretation.
Second, Gaynor et al. (2011) focus on a case where the
credit risk effect appears in the current period, while the
underlying economic reason may affect income in a later,
unreported period. In their experiment, ?rm pro?tability
22 M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38
re?ects only the credit risk effect, not the underlying
economic cause. Yet, prior studies showthat a ?rm’s perfor-
mance and its credit risk are intertwined (Ashbaugh-Skaife,
Collins, & LaFond, 2006; Barth, Hodder, et al., 2008; Bhojraj
& Sengupta, 2003). An increase in a ?rm’s credit risk may,
therefore, co-occur with a decrease in its pro?tability.
Because investors do not always treat pro?ts and losses
symmetrically (Bhojraj & Sengupta, 2003), Gaynor et al.’s
(2011) ?ndings for ?rms that report a pro?t might not
generalize to losses, particularly if pro?ts and losses have
different effects onthe perceptionof credit risk information.
Our study considers both pro?t and loss ?rms and thus sets
a boundary condition to Gaynor et al.’s (2011) results.
The rest of the paper proceeds as follows. In the next
section, we describe the background on the debate relating
to the presentation format of fair value gains and losses
due to credit risk changes. We also develop our hypothe-
ses. In the third section, we detail our experimental
procedure and, in the fourth, provide our results. In the
?nal section, we conclude and address the limitations of
our study.
Background and hypotheses development
The debate on reporting credit risk changes
In June 2009, the IASB staff prepared a paper summariz-
ing the pros and cons of including own credit risk in liabil-
ity measurement (IASB, 2009a). The IASB stated the
counterintuitive effect that a decrease in a ?rm’s credit-
worthiness increases net income as the main objection to
the fair value measurement of liabilities according to IAS
39. Comment letters largely con?rmed this view (IASB,
2009b). Respondents to the IASB discussion paper also
raised the concern that including credit risk effects in net
income increases its volatility (IASB, 2010). To mitigate
these problems, the IASB proposed IFRS 9, requiring the
credit risk effect to be included directly in OCI and the
remaining net change in the fair value to be reported in
pro?t or loss (IFRS 9.BC5.45).
Because the income effect of a ?rm’s credit risk changes
has been found to be counterintuitive (Gaynor et al., 2011;
Lipe, 2002), the discussion has focused mainly on whether
knowledgeable ?nancial statement users correctly evalu-
ate a credit risk gain as a negative signal and a loss as a
positive signal. Given the IASB’s proposal, the question
arises as to whether and how the new presentation format
would affect ?nancial statement users’ perceptions of ?rm
performance. We thus investigate the effects of reporting
credit risk changes in the OCI statement. Because previous
research ?nds similar results regarding the effects of credit
risk downgrades and upgrades in the absence of additional
disclosures (Gaynor et al., 2011), we limit our analysis to
the case of downgrades, that is, increases in credit risk.
A general framework for analyzing information processing
Based on the work of Hogarth (1980), Maines and
McDaniel (2000) provide a framework for modeling infor-
mation processing where they disaggregate information
processing into three phases: acquisition, evaluation, and
weighting. During acquisition, a person reads and stores a
piece of information in memory. Only after acquisition
can the reader incorporate the information into an overall
judgment. In addition, the reader must correctly interpret
the information (evaluation). During weighting, the last
step, the reader considers the importance of a particular
piece of information for the pertinent decision.
This framework has been adopted not only to the
reporting of credit risk (Lachmann, Wöhrmann, &
Wömpener, 2011) but also to various accounting contexts
(Elliott, Hodge, Kennedy, & Pronk, 2007; Hirst, Hopkins, &
Wahlen, 2004; Hodge, Kennedy, & Maines, 2004; Nelson
& Tayler, 2007). It also has been applied to various types
of decision makers, including nonprofessional investors
(Maines & McDaniel, 2000) and professional lenders (e.g.,
Hales, Venkataraman, & Wilks, 2011).
Characteristics of the presentation formats expected to
in?uence individual information processing
Credit risk gains are reported as components of net
income according to SFAS 159/IAS 39 and as part of OCI
under IFRS 9. We compare the presentation formats by
using two criteria: (1) the degree of isolation of credit risk
gains and (2) the nature of income of credit risk gains.
1
Building on the similarities and differences between the pre-
sentation formats concerning these criteria, we then develop
our hypotheses for the effects of presentation format in the
phases of information processing.
The ?rst dimension that differs across presentation for-
mats is the degree of isolation of credit risk gains. With an
increased amount of irrelevant information, i.e., with a
lower degree of isolation, it becomes harder to extract
the relevant information (Maines & McDaniel, 2000). Items
presented in the net income statement usually outnumber
those in the statement of comprehensive income.
2
Hence,
the degree of isolation for a single item is lower when
presented in the net income statement and higher when
presented in the OCI statement.
The second criterion for comparison is the nature of
income. Financial statement users will perceive credit risk
gains presented in the net income or the OCI statement dif-
ferently if they regard net income and OCI as elements that
address different kinds of information needs. Thus, equiva-
lent information might be seen differently if users with
limited attention focus on information that is part of a con-
struct considered more important for the question at hand
(Hirshleifer & Teoh, 2003). However, evidence on different
perceptions of the relative importance of net income
versus OCI is scarce. Studies that compare the two income
1
Research on information processing shows that labels (i.e., the deno-
tation of fair value changes as a ‘‘gain’’ or a ‘‘loss’’) can also affect
information processing. However, the presentation formats we analyze do
not differ with respect to labeling.
2
Our analysis of the 2010 annual reports of the Dow Jones Eurostoxx 50
companies con?rms this reasoning: On average, the net income statement
consists of 13.7 items, whereas the OCI statement comprises only 5.7 items
(excluding subtotals). Moreover, the net income statement, on average,
contains 4.7 (sub)totals, whereas the OCI statement has only 1.5
(sub)totals.
M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38 23
concepts usually dwell on the value relevance of special
items included in net income versus OCI gains and losses.
Jones and Smith (2011), for example, show that, although
OCI gains and losses are value relevant, their relevance is
signi?cantly lower than that of special items gains and
losses presented in net income. We therefore conclude that
the relevance of OCI for evaluating ?rm performance is
lower than that of net income.
Importance of net income as contextual information
Firm performance and changes in own credit risk are
closely related (Jiang, 2008). Further, ?rm pro?tability,
which is usually measured by net income (Hodder,
Koonce, & McAnally, 2001; Hopkins, 1996), is an important
contextual dimension that may be relevant for interpreting
credit risk changes. Thus, if ?nancial statement users are
particularly sensitive to changes in creditworthiness due
to changes in net income, differences in the two criteria
between the presentation formats can have a different
impact on information processing.
In this study, we compare information processing
between the net income and the OCI presentation formats
for situations in which the ?rm reports a pro?t and for sit-
uations in which a loss is realized. Prior experimental
research has sidestepped the moderating effects of changes
in a ?rm’s economic and ?nancial conditions by assuming
that the change in credit risk materializes before the
change in the ?rm’s economic condition does. In Gaynor
et al.’s (2011) experiment, for example, the fair value of lia-
bilities is affected in the current period, but the reason for
the credit risk change (i.e., the signing or the loss of a long-
term contract) in?uences income only in the following
(unreported) period. In other words, current-period pro?t-
ability does not re?ect the economic cause of the credit
risk change. This situation precludes a diagnosis of the rea-
sons that led to the credit risk increase, which is essential
for causal reasoning (Koonce, Seybert, & Smith, 2011).
Whether an increase in credit risk materializes before a
?rm’s economic condition deteriorates is a questionable
assumption. It is equally likely that rating downgrades
coincide with (or promptly follow) the worsening of a
?rm’s economic condition. Timing matters here. If inves-
tors are aware of the deterioration of a ?rm’s economic
condition, they may read the credit risk information more
diligently. Investors certainly assume an extremely nega-
tive change in the economic condition when net income
turns from positive to negative (e.g., Barth, Landsman, &
Lang, 2008; Burgstahler & Dichev, 1997; Leuz, Nanda, &
Wysocki, 2003).
Hypotheses development
Acquisition of credit risk information
The ?rst step in the information-processing framework
presented above is acquisition, which not only requires
reading but also storing and recalling information (Elliott
et al., 2007). Once the information is read, it must then
be represented ‘‘internally in a way that enables its use’’
(Hirshleifer & Teoh, 2003, p. 342). Internal representation
is affected by cognitive costs (Hirshleifer & Teoh, 2003)
and an item’s salience (Frederickson & Miller, 2004).
Hodge et al. (2004) ?nd that presentation format affects
acquisition. Because the degree of isolation is higher in
the OCI presentation format and a higher degree of isola-
tion leads to lower cognitive costs (Maines & McDaniel,
2000), individuals are more likely to acquire information
on credit risk gains if it is presented in the OCI rather than
in the net income statement. Consequently, we posit the
following hypothesis:
H1a. Individuals are more likely to acquire information on
credit risk changes when gains from credit risk increases
are presented in the OCI statement rather than in the net
income statement.
The earnings management literature shows that man-
agers try to avoid reporting small losses (Burgstahler &
Chuk, 2013; Burgstahler & Dichev, 1997; Jiang, 2008)
because losses are interpreted as ‘‘red ?ags’’. When ?nan-
cial statement users interpret a negative net income as a
warning, they might be more likely to read the ?nancial
information more diligently and to dedicate more atten-
tion to the components of net income to identify the reasons
for the loss. This will have a positive effect on the acquisi-
tion of credit risk changes when credit risk gains are
reported in the net income statement (where the loss is
reported) whereas acquisition is unaffected when credit
risk gains are reported in the OCI statement. In other
words, when a loss is reported in the income statement,
the likelihood of acquiring information on credit risk
changes increases under the net income presentation for-
mat but not under the OCI presentation format. Hence,
the relative advantage of OCI presentation, which stems
from its higher degree of isolation of information on credit
risk gains (H1a), should be reduced when a loss is reported.
However, we do not expect that the differences between
the presentation formats are completely eliminated when
a loss is reported because the net income statement
contains more items than the OCI statement, i.e., acquisi-
tion under the net income presentation format is still more
costly than under the OCI presentation format. We
therefore posit the following hypothesis:
H1b. Differences in the likelihood of acquiring information
on credit risk changes presented in the OCI statement
versus in the net income statement are reduced when a
loss is reported.
Panel A of Fig. 1 graphically depicts our predictions for
acquiring information on credit risk changes (H1a and
H1b).
Evaluation and weighting of information about credit risk
changes
Once ?nancial statement users acquire information on
credit risk gains, they evaluate and weigh it. Gaynor et al.
(2011) argue that users typically interpret gains as positive
and losses as negative signals (Hodder et al., 2001). Yet,
this intuition is misleading in the case of fair value adjust-
ments resulting from changes in own credit risk. However,
a credit risk increase leads to a gain under both presenta-
tion formats; we therefore expect no differences in the risk
24 M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38
of misinterpretation (i.e., evaluation) between the
presentation formats.
Presentation format can affect weighting in two
different and opposite ways. On the one hand, H1a states
that—because of the lower degree of isolation in the net
income presentation format—users are less likely to
acquire a credit risk change when credit risk gains are
reported in the net income statement. The degree of isola-
tion also affects the weighting phase: An investor who
does not acquire the credit risk change will place a low
weight on credit risk information when assessing ?rm
performance. Put differently, the lower degree of isolation
in the net income presentation format undermines
weighting (Nisbett, Zukier, & Lemley, 1981).
Conversely, the nature of income criterion suggests that
the user’s perception of the value relevance of credit risk
gains would be higher when these gains were presented
in the net income statement. Net income items are
regarded as better predictors of ?rm performance than
OCI items (Hodder et al., 2001; Hopkins, 1996; Jones &
Smith, 2011). Nonprofessional investors with limited cog-
nitive abilities or limited time will therefore focus more
on information contained in the net income statement
when assessing ?rm performance. Maines and McDaniel
(2000) also argue that a closer link to net income increases
the weight that nonprofessionals give to an income
component. Thus, the nature of income criterion results
in a positive effect on weighting when credit risk gains
appear in the net income statement.
Because the degree of isolation and the nature of
income pull in opposite directions, the overall effect of
the presentation format on weighting is unclear. To derive
a directional hypothesis, we focus only on those users who
have managed to acquire the change in credit risk and are
therefore not negatively affected by the lower degree of
isolation in the net income presentation format. The
weight they place on credit risk information is thus deter-
mined only by the nature of income. We predict that users
who have successfully acquired the change in credit risk
will place more weight on the corresponding information
when it is presented in the net income statement than
when presented in the OCI statement.
H2a. Presenting credit risk gains in the OCI statement,
rather than in the net income statement, decreases the
weight individuals place on previously acquired credit risk
information when assessing overall ?rm performance.
If the ?rm reports a loss, investors pay more attention
to identifying the reasons for the loss and less attention
to analyzing the gains that result from changes in the fair
value of liabilities. This shift in attention reduces the
weight placed on credit risk information. We expect that
the weights placed on credit risk gains are reduced to a lar-
ger extent when credit risk gains are presented in the net
income statement than when presented in the OCI
statement (where we expect no or only a small change in
the weights because the weights under the OCI presenta-
tion format are predicted to be already comparatively
low). Thus, the shift in attention due to the presence of a
loss should reduce – although not eliminate – the
differences in weighting between the presentation
formats. This leads to our next hypothesis.
H2b. Differences in the weighting of previously acquired
credit risk information presented in the OCI statement
versus in the net income statement are reduced when a
loss is reported.
Panel B of Fig. 1 graphically depicts our predictions for
weighting credit risk gains (H2a and H2b).
Overall evaluation of ?rm performance
In releasing IFRS 9, the IASB aimed to reduce net income
volatility and the likelihood of ?nancial statement users
misinterpreting ?rm performance (IASB, 2010). The IASB
did not argue that incorporating credit risk gains in OCI
would reduce credit risk misinterpretation but rather that
it would reduce ?rm performance misinterpretation.
Whereas the total amount of comprehensive income does
not depend on the presentation format, net income is
higher if it contains the credit risk gain (and OCI is lower
Panel A: Predicted pattern for H1
Panel B: Predicted pattern for H2
Panel C: Predicted pattern for H3
Fig. 1. Predictions for H1, H2, and H3.
M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38 25
by the same amount). Hence, all else equal, the presenta-
tion format leads to differences in the values of net income
and OCI.
The correct evaluation of ?rmperformance requires that
the investor mathematically subtract (i.e., neutralize) the
credit risk gain from income to calculate an ‘‘as if’’ income
that is not in?ated by the gain. Without this adjustment,
the investor’s estimate of ?rm performance is biased
upwards; that is, ?rm performance is misinterpreted. We
predict that the acquisition, evaluation, and weighting of
credit risk information determine whether the investor is
likely to neutralize the credit risk gain. We identify two
alternative cases in which an investor does not neutralize a
credit risk gain and misinterprets ?rm performance. As
explained below, we predict that both cases are more likely
whencredit risk gains are presentedinthe net income state-
ment. We do not require both arguments to simultaneously
hold for a speci?c investor, that is, we do not aggregate both
factors. Instead, we predict that each factor can separately
lead to the misinterpretation of ?rm performance.
First, some investors may not have acquired the credit
risk effect on income; these investors cannot neutralize
the gain because they are not aware of the relevant infor-
mation. When evaluating ?rm performance, they will rely
instead on the reported income measure, in?ated by the
fair value gain. H1a predicts that the acquisition of credit
risk changes is more likely when credit risk gains are
presented in the OCI statement. This should enable more
investors to neutralize the credit risk income effect.
Misinterpretation of ?rm performance should thus be less
likely in the OCI presentation format.
Second, some of those investors who have acquired the
credit risk information (and evaluated and weighted the
credit risk gain) still will not neutralize its effect. This
behavior is likely to occur for those investors who miscon-
strue the gain as a positive signal and consider it as highly
relevant for the evaluation of ?rm performance. We do not
make a prediction regarding differences in the risk of mis-
interpreting credit risk gains between the two presentation
formats, but we expect differences in weighting: H2a pre-
dicts that users will place less weight on the acquired
information on credit risk changes if the gains are pre-
sented as an OCI item (Hodder et al., 2001; Hopkins,
1996). If, in contrast, net income is used as a primary
source for the assessment of ?rm performance, then inves-
tors will place more weight on (potentially misinterpreted)
credit risk gains because net income is seen as a key pre-
dictor of the ?rm’s future development. We predict there-
fore that, when credit risk gains are reported in the net
income statement, fewer investors will neutralize the
credit risk income effect, and that investors will be more
likely to misinterpret ?rm performance.
In summary, we offer two mechanisms by which an
investor is less likely to neutralize a credit risk gain when
credit risk gains are presented in the net income than in
the OCI statement. We make our formal prediction below.
H3a. Individuals are less likely to make biased estimates of
overall ?rm performance when credit risk gains are
presented in the OCI statement rather than in the net
income statement.
If net income turns negative, then the change in a ?rm’s
economic condition becomes obvious. Even if users regard
the credit risk gain as a positive signal, the overall effect on
net income is negative. Therefore, we hypothesize that the
difference between the presentation formats regarding
overall ?rm performance evaluation is reduced in the pres-
ence of a loss. We expect that this decrease in the differ-
ence is mainly driven by less biased ?rm performance
evaluations under the net income presentation format.
For the OCI presentation format, in contrast, we expect
no (or only a small) change regarding the likelihood of
biased ?rm performance evaluations. The reasons are that
net income is not in?ated by the credit risk gain in the OCI
presentation format, and that OCI is seen as less predictive
for the ?rm’s future development than net income. Since
investors rely on net income for their assessment of ?rm
performance and net income is higher by the amount of
the fair value gain in the net income presentation format,
we do not expect that the differences in ?rm performance
evaluation between the presentation formats are
completely eliminated when a loss is reported.
H3b. Differences in the likelihood of making biased esti-
mates of overall ?rm performance between the presenta-
tion formats are reduced when a loss is reported.
Panel C of Fig. 1 graphically depicts our predictions for
?rm performance evaluation (H3a and H3b). The lower
part of Fig. 2 shows how the information-processing
framework relates to our hypotheses.
Experimental method
Design
We conducted an experiment using a 2 Â 2 between-
subjects full-factorial design. The ?rst treatment variable
was the presentation format of a fair value gain resulting
from an increase in credit risk. This gain was either
included in the net income statement (NI), as required by
SFAS No. 159 and IAS 39, or in the other comprehensive
income statement (OCI), as required by IFRS 9. Under both
conditions, the credit risk increase led to a gain of
€1,920,000. The notes informed participants that the fair
value gain was attributable entirely to a change in the
?rm’s credit risk. However, the notes did not provide infor-
mation on the direction of the credit risk change.
3
The second treatment variable was whether a negative
pro?tability development (LOSS) occurred simultaneously
with the downgrade or not (PROFIT). Asset impairments
are a common reason for net income to turn negative when
economic conditions deteriorate (Barth, Landsman, et al.,
2008). In the LOSS treatment, an R&D impairment of
€3,900,000 overcompensated for the positive income effect
from the fair value gain. Thus, both net income and
3
Neither GAAP nor IFRS require disclosure of an explanation regarding
the direction of the credit risk change. Thus, preparers usually do not
provide such disclosures. We analyzed the 2011 annual reports of the 10
largest European and 10 largest US banks. We did not ?nd even one
institution that disclosed the direction of a credit risk change when
providing information on the calculation of credit risk gains/losses.
26 M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38
comprehensive income were negative in the LOSS treat-
ment and positive in the PROFIT treatment. All remaining
?nancial items were identical for the LOSS and PROFIT
conditions.
Participants
We used 93 auditors as participants. This allows us to
compare our results to the ?ndings of Gaynor et al.
(2011) who use CPAs. We regard auditors as an appropriate
proxy for knowledgeable nonprofessional investors
because they have a reasonable degree of ?nancial knowl-
edge, as required by the FASB and the IASB, to understand
?nancial statements (Concepts Statement No. 8, IASB
Conceptual Framework).
4
The professional activity of 88% of our participants was
auditing. The remaining 12% worked in the ?eld of audit-
related services. Ninety-?ve percent worked for audit
?rms. Thirty-two percent of these participants were
employed by one of the Big 4, and 68% by non-Big 4 audit
?rms. On average, our participants had 6.8 years of profes-
sional experience and were 34.3 years old.
5
They also had,
on average, 4.9 years of experience in making private invest-
ment decisions. We randomly assigned our participants to
the four treatment conditions NI_PROFIT (n = 22), NI_LOSS
(n = 24), OCI_PROFIT (n = 24), and OCI_LOSS (n = 23). We did
not detect any signi?cant differences between the four
treatment groups with respect to personal characteristics
(all p-values > 0.1).
Procedure and payments
We ran a paper-and-pencil experiment during a profes-
sional training seminar.
6
The experiment took 45 minutes
on average. We gave participants printed instructions con-
taining a brief description of the accounting rules. Partici-
pants under the NI condition learned from these rules that
gains and losses resulting from credit risk changes were pre-
sented in the net income statement, whereas participants
under the OCI condition learned that these gains and losses
were presented in the OCI statement. After reading the
instructions, participants received three sealed envelopes
at different times during the experiment. The envelopes con-
tained materials and question sets that we asked our partic-
ipants to complete consecutively.
The participants received a ?xed fee of €7 and a perfor-
mance-contingent payment between €0 and €10. During
the experiment, we rewarded them with the experimental
currency ‘‘Points’’. At the end of the experiment, we con-
verted Points into euros (100 Points = €1 US$1.3 at the
time of the experiment) and anonymously paid them in
cash. They earned 100 Points for a correct answer and lost
100 Points for a wrong answer.
Because risk preferences can affect perceptions of cred-
itworthiness, we elicited our participants’ risk preferences
Fig. 2. Hypotheses, related questions, and the information processing framework.
4
Though many studies use MBA students as a proxy for nonprofessional
investors (Elliott et al., 2007), some use CPAs, even though the experimen-
tal task is an investment task rather than an auditing task (e.g.,
Frederickson, Hodge, & Pratt, 2006; Gaynor et al., 2011; Ghani, Laswad, &
Tooley, 2011). Gaynor et al. (2011) argue that understanding credit risk
information is complex and that MBA students might not be sophisticated
enough to interpret this information properly.
5
Unlike in the United States, where two years of work experience is
usually required before becoming a CPA, in our case—barring minor
exceptions—graduates need three years and undergraduates need four
years of work experience before being admitted to the exam.
6
The experiment was conducted in Germany. A professional editing
service translated all the excerpts from the instrument that are contained in
this paper into English, and we double-checked the translation. Since the
IFRS are available both in English and in German, we used the of?cial
translation where possible.
M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38 27
using a lottery task (Dohmen et al., 2011; Sprinkle,
Williamson, & Upton, 2008). They had to choose, for 20
cases, between a lottery with equally probable opportuni-
ties to earn 300 Points or zero Points and a safe payment
s (0 6 s 6 190 Points). We followed Dohmen et al. (2011)
and classi?ed participants based on their choices as risk
averse, risk neutral, or risk taking and added the payoff
earned in the lottery task to the participants’ compensation.
We also included three problems from the Cognitive
Re?ection Test (CRT) suggested by Frederick (2005) to eli-
cit a speci?c cognitive ability. All of the questions from the
CRT provoke an intuitive but wrong answer, whereas the
correct answer requires more re?ection. The CRT problems
and our experimental task require similar abilities because
interpreting a credit risk gain as a positive signal is also
intuitive but incorrect. We did not provide monetary
incentives for completing the CRT.
Total compensation varied from €9 to €20, with an
average of €17.59. Participants in the NI_PROFIT/NI_LOSS/
OCI_PROFIT/OCI_LOSS conditions received €16.96/€18.02/
€17.40/€17.94. There was no signi?cant difference in
compensation among the four groups (Kruskal–Wallis test,
v
2
= 1.45, p = 0.69).
Task and instrument
Participants had ?ve minutes to read the instructions
and the accounting rules. They then received the ?nancial
statement of the ?ctitious Company XY with data for the
current and the preceding period (balance sheet, statement
of comprehensive income, cash ?ow statement, and foot-
note disclosures). Panel A of the appendix summarizes
the key ?nancial information. Re?ecting our observations
of business practice, the degree of isolation for credit risk
information was higher under the OCI than under the NI
treatment conditions. In the OCI treatments, the credit risk
gain was one of four entries included in the OCI statement,
while in the NI treatment it was one of eight.
Participants had eight minutes to read the ?nancial
statements before we handed out the ?rst envelope, con-
taining two questions. The ?rst question asked for an esti-
mate of the stock price change in the current period
(relative to the previous year).
7
Speci?cally, participants
had to estimate on a scale from À11 to +11 whether the
stock price had declined considerably (À11), remained
unchanged (0), or increased considerably (+11) in the cur-
rent year relative to the preceding year. If no credit risk gain
had occurred in the current period, then the deterioration of
?rm performance would have been obvious under all four
conditions. Investors who did not neutralize the credit risk
gain from income were misled to provide a biased, that is,
too positive stock price estimate. Participants’ stock price
estimates were used to test H3 (?rm performance evalua-
tion). The second question asked participants to rate the
relevance of ?ve factors (return on equity, cash ?ow, compo-
sition of earnings and expenses, creditworthiness, and capi-
tal structure) when evaluating the stock price of Company
XY by using a scale from 0 (not at all important) to 11 (extre-
mely important). We used the answers to this question to
test H2 (weighting).
We then collected the materials and the ?nancial state-
ments from participants before giving them a second set of
questions. This time, the ?rst question asked whether
Company XY had experienced a change in its creditworthi-
ness thus verifying whether our participants had acquired
the change in creditworthiness (H1) (acquisition). The sec-
ond question asked whether there had been a change in
the fair value of long-term debt. We used this question
as a supplemental test. We then collected these materials
back and handed out the last set of questions.
Participants could again refer to the ?nancial state-
ments when completing this third set of questions. Their
answers to these additional questions were not used for
testing our hypotheses but to better understand their
behavior. First, we asked them to imagine holding a portfo-
lio of shares of ten different companies, including shares of
Company XY. We then instructed them to buy additional
shares if the ?rm’s creditworthiness had increased and to
sell all of the shares if the ?rm’s creditworthiness had
decreased. We asked which option (buy or sell) they would
choose if their decision exclusively depended on the
change in the ?rm’s creditworthiness. Following Gaynor
et al. (2011), we used this question as a manipulation
check to determine whether participants correctly inter-
preted the credit risk change (evaluation of creditworthi-
ness). We asked them to brie?y explain their decision in
writing. As a supplemental test, we requested an estimate
of the magnitude of the change in creditworthiness on a
scale from 0 (no change) to 11 (very strong change).
Fig. 2 presents the relations between the questions, our
hypotheses, and the phases in the information-processing
framework.
Finally, we applied the lottery task and the CRT. At the
end of the experiment, participants answered manipula-
tion check questions and provided demographic data.
Results
Manipulation and other checks
The ?rst manipulation check question asked whether
fair value gains and losses occurring from changes in
own credit risk were included in net income or in OCI. In
both the OCI and NI treatments, 95.7% of our participants
responded correctly. Next, we asked whether there was
an impairment of an R&D project in the current period.
For the PROFIT (LOSS) treatments, 91.3% (100%) of our par-
ticipants answered this question correctly. These results
indicate that both manipulations were successful.
8
7
Since participants had to answer the questions concerning a ?ctitious
company and we wanted to rule out assumptions about future develop-
ments, we asked for an estimate of the stock price of the current period
relative to the previous period. If we had asked for the stock price of a future
period, participants might have made different assumptions concerning the
predictive power of the (transitory) loss under the LOSS conditions.
8
The results presented in the following sections are based on the full
sample. Unless stated otherwise, the results are identical if we exclude
participants who responded incorrectly to the manipulation check
questions.
28 M. Lachmann et al. / Accounting, Organizations and Society 41 (2015) 21–38
As stated above, we inferred our participants’ interpre-
tations of the credit risk gain by informing them that they
should sell all of their shares of the ?ctitious ?rm when
credit risk had increased. We take the participants’ deci-
sions not to sell the shares as an indication of a misinter-
pretation of the credit risk gain. To verify that our
participants correctly understood this question, we asked
them the same question for a different ?rm, Company Z,
but told them that the credit risk of this ?rmhad increased.
Some 97% of our participants correctly answered that they
should sell the shares of Company Z.
We elicited risk preferences using the lottery task but
could not classify 17% of our participants because they pro-
vided inconsistent responses across the different lotteries.
Of the remaining, we found that 43% were risk neutral, 43%
were risk averse, and 14% were risk taking. We also mea-
sured their cognitive abilities by using the number of cor-
rect answers to the CRT questions, that is, the CRT score.
Because some of the participants did not answer the sec-
ond and/or third question of the CRT (17% and 25%, respec-
tively), we decided to consider only the ?rst CRT question,
which only one participant did not answer.
9
A total of 40%
of participants answered the ?rst question correctly. The
results from the lottery task and the ?rst question of the
CRT show no statistically signi?cant differences regarding
risk aversion or cognitive abilities between the four treat-
ment groups (p-values > 0.1).
10
Test of H1 (acquisition)
Hypothesis 1a states that the likelihood of acquiring
credit risk information is higher when credit risk gains
are reported in OCI. As presented in Panel A of Table 1,
we ?nd that, on average, 74% of our participants under
the NI conditions and 89% under the OCI conditions cor-
rectly recalled the change in creditworthiness. In line with
our prediction in H1a, we ?nd that the likelihood of acquir-
ing credit risk information is higher when credit risk gains
are reported in OCI (v
2
= 4.38, p = 0.04), as depicted in
Panel B of Table 1. Panel A of Fig. 3 compares our predic-
tions and our results for H1.
When the ?ctitious ?rm reported a pro?t, 59% (88%) in
the NI_PROFIT (OCI_PROFIT) cell acquired the credit risk
change (i.e., the difference across the NI and the OCI treat-
ments in the percentage of participants acquiring this
information is 29%). This difference is signi?cant (chi-
squared-test, v
2
= 4.80, p = 0.03, two-tailed). Yet, when
the ?ctitious ?rm reported a loss, 88% (91%) of the partic-
ipants in the NI_LOSS (OCI_LOSS) cell acquired the credit
risk change; i.e., the difference in acquisition falls to 3%
and becomes insigni?cant (Fisher’s exact test, p = 1.00,
two-tailed). This ?nding is in line with H1b that posits that
differences in the likelihood of acquiring information on
credit risk changes between the presentation formats are
reduced when a loss is reported. However, the ANOVA
interaction term NI Â LOSS in Panel B of Table 1 is insignif-
icant (v
2
= 0.92, p = 0.34). To further analyze the functional
form of the interaction, we test H1b using planned
contrasts, which has two advantages (Buckless &
Ravenscroft, 1990). First, contrast codes increase the statis-
tical power without increasing Type I error rates. Second,
we predict an ordinal interaction. Consistent with our pre-
diction, we use contrast weights of À3 for the NI _PROFIT
condition, À1 for the NI _LOSS condition, and +2 for the
OCI conditions (i.e., OCI_PROFIT and OCI_LOSS) (see Fig. 3,
Table 1
Acquisition (H1).
Pro?tability Presentation format
P
NI OCI
Panel A: Descriptive statistics
Dependent variable: Change in credit risk acquired (binary)
PROFIT 59% (n = 22) 88% (n = 24) 74%
LOSS 88% (n = 24) 91% (n = 23) 89%
P
74% 89% 82%
Source df v
2
p-Value
Panel B: Categorical modeling statistics (logistic regression)—test of H1a (n = 93)
Dependent variable: Change in credit risk acquired (binary)
Intercept 1 9.94