Description
Weconductalaboratoryexperimenttoexamineinvestors’assessmentsofmanagers’stewardship.Weprovideevidencethatinvestorstendtoattributeexternal(i.e.,non-manager-related)causesoffirmper-formancetomanagers’performance.Wepredictandfindthatfairvalueinformationenablesinvestorstoovercomethistendencyandmakebetterstewardshipdecisionsthaninvestorswitamortizedcostinformation.Wealsofindthatinvestorspresentedwithamortized-cost-basedfinancialstatementsper-formbettertotheextenttheyaccessfair-value-basedfootnoteinformation,whileinvestorspresentedwithfair-value-basedfinancialstatementsperformworsetotheextenttheyaccessamortized-cost-basedfootnoteinformation.
Accounting, Organizations and Society 46 (2015) 100–114
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
The effect of alternative accounting measurement bases on investors’
assessments of managers’ stewardship
Spencer B. Anderson, Jason L. Brown, Leslie Hodder, Patrick E. Hopkins
?
Kelley School of Business, Indiana University, 1309 E. Tenth Street, Bloomington, IN 47405, United States
a r t i c l e i n f o
Article history:
Received 1 June 2014
Revised 30 March 2015
Accepted 31 March 2015
Available online 29 April 2015
Keywords:
Stewardship
Fair value
Experimental market
Transparency
Comparability
Attribution theory
a b s t r a c t
We conduct a laboratory experiment to examine investors’ assessments of managers’ stewardship. We
provide evidence that investors tend to attribute external (i.e., non-manager-related) causes of ?rm per-
formance to managers’ performance. We predict and ?nd that fair value information enables investors
to overcome this tendency and make better stewardship decisions than investors with amortized cost
information. We also ?nd that investors presented with amortized-cost-based ?nancial statements per-
form better to the extent they access fair-value-based footnote information, while investors presented
with fair-value-based ?nancial statements perform worse to the extent they access amortized-cost-based
footnote information. Collectively, our results suggest that investors’ stewardship decisions are improved
because fair value information more transparently provides the information required to properly consider
the opportunity costs associated with managers’ actions and disentangle endogenous actions by managers
from exogenous market forces that are outside of managers’ control.
© 2015 Elsevier Ltd. All rights reserved.
Introduction
We investigate whether, when holding information constant,
the use of fair value measurements in the primary ?nancial state-
ments improves ?nancial statement users’ (hereafter, “investors’”)
judgments about managers’ stewardship. Opponents of incorporat-
ing fair value measurements into the primary ?nancial statements
argue that fair values do not meet information needs relative to
the stewardship objective of ?nancial reporting, and maintain that
amortized-cost-based ?nancial statements are necessary to assess
stewardship (e.g., Holthausen & Watts, 2001; Nissim & Penman,
2008). We propose that, compared to amortized-cost-based mea-
surement, reported fair value measures in the primary ?nancial
statements provide information in a more transparent manner that
will reduce the burden of processing and improve the accuracy of
stewardship assessments.
As noted by the IASB (2005) staff, the word “stewardship”
has many different meanings in the ?nancial accounting standard-
setting literature. Consistent with the root word “steward,” we pro-
pose that the focus of stewardship is the assessment of actions
taken by stewards (i.e., managers) in the discharge of their respon-
?
Corresponding author.
E-mail addresses: [email protected] (S.B. Anderson), [email protected]
(J.L. Brown), [email protected] (L. Hodder), [email protected] (P.E. Hopkins).
sibilities. In the modern corporation, these responsibilities relate
to an agency relationship and the performance of some delegated
activities on behalf of another party. These delegated activities in-
clude not only safeguarding of assets but also administering busi-
ness resources in a way that generates an acceptable rate of return
on those resources. Therefore, we de?ne stewardship as the actions
of managers in the discharge of their responsibilities to preserve
the value of investors’ capital investment and to earn a commen-
surate return on that investment.
Standard setters currently offer differing positions on the
prominence of stewardship within the overall objective of ?nan-
cial reporting. The FASB recently consolidated the objective of ?-
nancial reporting to resource allocation, subordinating the steward-
ship objective in the process, while stating that information that
is bene?cial for resource allocation is also bene?cial for steward-
ship assessment (FASB, 2010a). Meanwhile, the IASB recently ten-
tatively decided to amend the Conceptual Framework to increase
the prominence of stewardship within the overall objectives of ?-
nancial reporting (IASB, 2014). Our study provides empirical evi-
dence relevant to the stewardship and ?nancial reporting debate,
and addresses O’Connell’s (2007) observation that “stewardship-
related questions have received insu?cient attention in the ?nan-
cial reporting literature to date—particularly from an empirical per-
spective” (p. 216).http://dx.doi.org/10.1016/j.aos.2015.03.007
0361-3682/© 2015 Elsevier Ltd. All rights reserved.
S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114 101
In addition to the controversy over ?nancial reporting objec-
tives, the question of which is the most appropriate measurement
basis for recognized ?nancial statement elements continues to be
one of the most di?cult issues addressed by standard setters. As
either a cause or a consequence of this di?culty, the conceptual
frameworks of the FASB and IASB presently are largely devoid
of any conceptual guidance on accounting measurement issues.
Speci?cally, the FASB Statement of Financial Accounting Concepts
No. 5 (SFAC 5), Recognition and Measurement in Financial Statements
of Business Enterprises (1984, par. 66–70) and the IASB’s Concep-
tual Framework (2014, par. 4.54–4.56) include only terse lists of
measurement bases historically observed in ?nancial statements,
and provide no guidance about the conditions under which one
measurement basis may be preferred over another. According to
Whittington (2008, p. 154), continuing work on the measurement
sections of the conceptual frameworks is likely to be lengthy and
highly contentious because measurement issues often are reduced
to an unproductive competition between fair value and amortized
cost bases of measurement.
1
To correctly assess stewardship, we propose that investors must
(1) receive information that is su?ciently relevant and su?ciently
precise for the task, and (2) disentangle management’s contribu-
tion to ?rm performance from factors outside of management con-
trol (e.g., environmental factors).
2
In this study, we leverage the
comparative advantage of controlled experiments to hold constant
the ?rst requirement to allow us to evaluate the effect of mea-
surement basis on the second. Speci?cally, we conduct a laboratory
experiment to provide empirical evidence relevant to the assess-
ments of stewardship and measurement bases (i.e., amortized cost
and fair value). Our tests are motivated by analytic and psychology
research. Analytic research suggests that high-quality stewardship
assessments depend on principals’ ability to separate the effects of
manager actions from the effects of non-manager (i.e., exogenous)
conditions and events that affect the ?rm (Harris & Raviv, 1978;
Holmström, 1979; Stiglitz, 1974). However, research on attribution
in psychology suggests that investors will have di?culty accurately
assessing managers’ contribution to ?rm performance because in-
dividuals generally have a hard-wired tendency to attribute cause
to individual actors rather than accurately identifying the impact
of exogenous or circumstantial factors (Ross, 1977).
Theory suggests that these tendencies persist because observers
attribute outcomes to salient features of the setting and usually
fail to make the necessary adjustments to correct the automatic
perceptual process (Gilbert, Krull, & Pelham, 1988; Taylor & Fiske,
1975). By making information about exogenous situational factors
more salient, this tendency can be mitigated because less effort
is required to consider situational factors (Gilbert, 1989; Quat-
trone, 1982). We propose that, compared to amortized cost, fair
value measurements provide more transparent information that
can more easily reveal the effects of managerial actions and ex-
ogenous non-manager conditions and events.
Although our primary focus is on measurements recognized in
the primary ?nancial statements, we also investigate the in?u-
ence of notes to the ?nancial statements that include information
about the other measurement bases on investors’ stewardship as-
sessments and investment decisions.
3
Because fair value informa-
1
The debate about the use of fair values in ?nancial statements has been on-
going for most of the 20th Century and has crept into the 21st Century (e.g.,
Chambers, 1966; MacNeal, 1939; Paton & Littleton, 1940; Watts, 2003). As described
by Hodder, Hopkins, and Schipper (2014), many of the arguments on both sides of
the debate remain seemingly intact and untouched by the passage of time.
2
As discussed later in the manuscript, this is consistent with the agency litera-
ture. See Baiman (1982), Hart and Holmström (1987) and Baiman (1990) for com-
prehensive reviews of this literature.
3
Supplemental information about alternative measurement bases is fairly com-
mon in published ?nancial statements (e.g., Statement of Financial Accounting Stan-
tion should be perceived as more transparent, we propose that in-
vestors using amortized-cost-based ?nancial statements will make
higher quality stewardship assessments if they take the time and
effort to process the fair value information in the notes. However,
because amortized cost information provides less salient signals
about exogenous shocks, we expect lower-quality stewardship as-
sessments and investment decisions for users who view fair-value-
based ?nancial statements and spend additional time and effort
processing the amortized-cost-based information in the footnotes
(Hackenbrack, 1992; Nisbett, Zukier, & Lemley, 1981).
In our experiment, participants assume the role of investors
who assess the quality of the ?rm’s management and/or allocate
investment capital to the ?rm. A controlled experiment has ad-
vantages for testing our research questions because we are able
to investigate the effects of a fair value recognition regime that
is largely unobservable in naturally occurring settings and because
we are able to collect information on investors’ decision processes
and judgments that are unavailable in archival databases. Further,
our setting allows us to carefully structure an economic environ-
ment where changes in ?rm value are determined solely by a
combination of (1) exogenous, systematic market forces and (2)
managerial actions conditional on those forces. Ensuring that fun-
damental values and reported results are determined only by a
combination of exogenous shocks and managerial actions provides
greater con?dence that the differences we obtain across measure-
ment bases are strictly a function of investors’ ability to isolate the
contribution of managerial actions to ?rm value.
Participants in the experiment receive information in the form
of ?nancial statements. Approximately half of the participants re-
ceive ?nancial statements in which assets and income are based on
fair value (amortized cost). We also provide to all participants foot-
note information that includes asset and income information using
the measurement basis that is not included on the face of the ?-
nancial statements (i.e., the total information set is the same across
all participants). Speci?cally, participants in the fair-value (amor-
tized cost) conditions can refer to footnote disclosures to obtain
amortized cost (fair value) asset information.
Our results support our prediction that, holding information
constant, recognition of fair values in the primary ?nancial state-
ments enables investors to more accurately assess the quality of
managers’ actions. This is due, at least in part, to the ability of
fair value to ameliorate the tendency to attribute exogenous forces
to management that is apparent when investors view amortized
cost ?nancial statements. Speci?cally, we ?nd that when a nega-
tive (positive) exogenous shock affects ?rm value, investors view-
ing ?nancial statements that report amortized cost measurement
are more likely to believe that ?rm management’s decisions were
of relatively lower (higher) quality, regardless of actual manage-
ment decision quality. For investors viewing fair-value-based ?nan-
cial statements, this bias is mitigated.
By examining process data, we con?rm that better stewardship
assessments and resource allocation decisions are attributable to
investors’ greater (lesser) use of fair value (amortized cost) infor-
mation, even when that information is in the footnotes. Speci?-
cally, investors in the amortized cost condition spending more time
with fair value footnotes performed better relative to investors
spending less time on the fair value footnotes. In contrast, in-
vestors in the fair value condition spending more time on amor-
tized cost footnotes performed worse than investors spending less
time on amortized cost footnotes.
dards No. 107, “Disclosures about Fair Value of Financial Instruments” [FASB 1991]
currently codi?ed in FASB Accounting Standards Codi?cation Topic 825) and in-
creased levels of such disclosure have been included in recent standard-setting pro-
posals (e.g., FASB, 2010b, par. 85).
102 S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114
We organize the rest of this paper as follows: In the next sec-
tion, we outline our theory and hypotheses; in the third section,
we describe our operationalization of stewardship, experimental
task, and research design; in the fourth section we discuss our re-
sults; and in the ?fth section we document our conclusions, de-
scribe important caveats to the generalizability of our study and
provide directions for further research.
Theory and hypotheses
Intersection of accounting measurement and stewardship
As noted by many accounting commentators, a theoretically
neutral starting place for evaluating the relative merits of alterna-
tive measurement bases in the primary ?nancial statements should
begin with identifying the primary objectives of ?nancial reporting.
Following the identi?cation of objectives, the usefulness of alter-
native measurement bases can be assessed relative to these objec-
tives. However, at least some of the literature con?ates these two
steps by holding self-evident the joint propositions that (1) stew-
ardship is the most important objective of ?nancial reporting and
(2) good stewardship can be achieved only with amortized-cost-
based ?nancial statements (e.g., Birnberg, 1980; Chen, 1975).
Why stewardship and amortized cost are so closely linked
is unclear; however, several authors assert that amortized cost,
transaction-based accounting information is desirable primarily
because it is more veri?able (e.g., Demski & Sappington, 1993;
Watts, 2003). Researchers have noted the linkages between stew-
ardship and agency theory (Baiman, 1990; Birnberg, 1980; Gjes-
dal, 1981). More recent research suggests that information rel-
evant for incentive and control purposes necessarily focuses on
completed transactions and past performance, rather than esti-
mates and future-oriented projections (Ball, 2001; Barclay, Gode,
& Kothari, 2005). These arguments shift the focus of the measure-
ment basis debate by introducing a more speci?c set of poten-
tially testable, but largely untested, assertions: (1) that amortized-
cost-based accounting information is inherently more veri?-
able than fair-value-based accounting information and (2) that
amortized-cost-based ?nancial information is inherently more con-
tractible than fair-value-based ?nancial information in an agency
setting.
There are reasons to question—and for scholars ultimately to
test—both of these assertions. First, despite persistent claims (e.g.,
Ramanna & Watts, 2012), research provides no empirical evidence
that fair values are inherently less veri?able than amortized costs
(see discussion in Hodder et al. (2014)). Although fair valuation po-
tentially introduces measurement uncertainty into ?nancial state-
ments, the practice of deferring and amortizing costs potentially
introduces both measurement uncertainty and existence uncer-
tainty. This is because amortized cost ?nancial statements are
based on a set of accruals, deferrals, and allocations, the objective
appropriateness of which is di?cult to independently verify.
4
As
Glover, Ijiri, Levine, and Liang (2005) acknowledge, ?nancial infor-
mation can be classi?ed along a continuum of veri?ability across
all measurement bases. Along this continuum, some fair-value-
based measurements are likely to be more veri?able than some
amortized-cost-based measurements. For example, the change in
fair value of a marketable ?nancial instrument may be much eas-
ier to verify than whether revenue has been earned in a particu-
lar period. Gjesdal (1981) recognizes that veri?ability is not abso-
lute, and establishes that information useful for stewardship must
4
For example, although an expenditure of cash may be easily veri?ed, there may
be substantial uncertainty about the existence of future economic bene?ts that may
be created by that expenditure, or whether the value of those bene?ts, if they exist,
is equal to the accumulated or allocated costs expended.
be only su?ciently veri?able at a reasonable cost. Because both
fair values and amortized costs may meet these criteria, veri?abil-
ity cannot be a su?cient justi?cation for rejecting fair value as a
measurement basis useful for stewardship.
Second, although evaluating managers based on realized ob-
servable outcomes facilitates “settling up” that may prevent ex post
inequitable incentive payments (Leone, Wu, & Zimmerman, 2006),
such contracts may not incent optimal ex ante actions, and these
ex ante actions more generally are the focus of contracting in an
agency setting. For example, if incentive compensation is based on
realizations, and outcome risk is su?ciently high, then risk-averse
managers may be unwilling to accept projects that clearly are in
the best interests of the shareholders (Weiss, 2011). In contrast
to the view that performance information should facilitate observ-
able outcome-based settling up, the agency literature establishes
that performance information is most useful for control when it
reduces uncertainty about agents’ current actions, not necessarily
current results (Gjesdal, 1981; Paul, 1992). If changes in fair val-
ues re?ect the present value of managerial actions, then fair values
may be useful for contracting precisely because they are not real-
ized outcomes, and instead re?ect forward-looking aggregations of
currently available information. If contracts based on “settling up”
are not necessarily more e?cient than contracts based on aggre-
gations of currently available information, the fact that fair values
re?ect forward-looking information cannot be a su?cient justi?-
cation for rejecting fair value measurement as a basis useful for
stewardship.
The fact that veri?ability and ex post settling-up value are
not su?cient reasons for rejecting fair value as a stewardship-
relevant measurement basis also does not imply that amortized
cost-basis measurements should be rejected in favor of fair value
measurements. We propose that identifying the conditions under
which fair-value-based or amortized-cost-based measurements are
more useful for stewardship is an empirical question. According
to agency theory, the answer to this question will be a function
of which measurement basis best helps investors isolate the ef-
fects of managerial actions on ?rm value. As Paul (1992) sum-
marizes, “to provide optimal incentives in a principal-agent prob-
lem, we need to weight information according to its informative-
ness about the manager’s contribution, or value-added, to the ?rm”
(p. 472).
Both amortized cost and fair value measurement bases produce
summary outputs that re?ect the effects of managerial actions;
however, both measures also confound the effects of managerial
actions with other exogenous events that impact the ?rm. For
this reason, we posit that, holding information constant, the rel-
ative usefulness of alternative measurement bases in the primary
?nancial statements must be a function of transparency. We de-
?ne transparency loosely as ?nancial statement presentation that
facilitates the extraction of decision-relevant information.
5
Speci?-
cally, even if the contribution of the manager theoretically can be
extracted from aggregated ?nancial results by appropriate use of
all recognized and disclosed information, investors may improp-
erly attribute variation in ?rm performance to variation in man-
ager actions if the effect of those actions is not su?ciently trans-
parent. This is consistent with psychology research on attribution,
which ?nds that individuals have a natural tendency to attribute
cause to individuals, rather than considering exogenous or circum-
stantial factors (Ross, 1977). We discuss this theory in the next
section.
5
Fair value measurement also has other attributes that may increase informative-
ness relative to amortized costs. For example, the greater timeliness of fair values
is well documented (see Laux and Leuz (2009) for a discussion).
S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114 103
Attribution theory
Incorrectly attributing changes in company performance to
manager actions undermines the e?ciency of incentive contracts
that are based on ?nancial results (Gjesdal, 1981; Paul, 1992). Sev-
eral models exist to describe the attribution process, but all gen-
erally agree that the process of attributing cause to events can be
broken into two steps: ?rst, attributions are initially automatically
activated based on observed stimuli, and second, more controlled
processes may override initial automatic judgments when activated
(Gilbert et al., 1988; Quattrone, 1982; Trope, 1986; Winter & Ule-
man, 1984). These models also posit that initial attributional judg-
ments on individuals can be incorrect, particularly when multiple
causes for behavior are at play and ambiguity shrouds the cause–
effect relationship.
Extant research on causal inference in psychology suggests that
when observers initially attribute causes for outcomes, they tend
to overweight personal dispositions in human actors (Ross, 1977).
That is, in an environment where an outcome is dependent on
two factors, one based on manager actions and the other resulting
from an exogenous in?uence, individuals will overweight manage-
ment’s role and underweight the role of exogenous factors. Consis-
tent with this theory, we predict that when a report of ?rm per-
formance is provided, investors will be prone to attribute cause to
managers, irrespective of managers’ contribution. This leads to our
?rst hypothesis:
H1a. In their manager quality assessments, investors will posi-
tively attribute exogenous (i.e., non-manager) components of ?rm
performance to managers.
Effect of measurement bases on attribution errors
To overcome the tendency of investors to incorrectly attribute
exogenous components of ?rm performance to managers’ actions,
investors must correctly identify the direction and magnitude of
managerial actions on ?rm performance, independent of the im-
pact of exogenous factors. However, because the tendency to at-
tribute cause for ?rm outcomes to managers is automatic, over-
coming this potential bias may be di?cult (Winter & Uleman,
1984). These tendencies that result in attribution errors are theo-
rized to persist because (1) observers attribute outcomes to poten-
tial causes that capture attention and (2) individuals fail to make a
deliberate and conscious effort to adjust original automatic causal
attributions (Gilbert et al., 1988; Taylor & Fiske, 1975).
One way to correct for investors’ tendency to attribute causes
of ?rm outcomes to managers is by increasing the transparency of
exogenous (i.e., non-manager-related) factors. In most settings, in-
dividuals’ attributions focus on individuals rather than situational
factors that are more di?cult to visualize (Lassiter, Geers, Munhall,
Ploutz-Snyder, & Breitenbecher, 2002). Therefore, when informa-
tion regarding situational factors at play is provided more transpar-
ently, systematic attribution errors can be corrected by (1) shift-
ing the perceptual focus of potential causes to exogenous forces
(Gilbert, 2002) and (2) decreasing the effort required to consider
situational factors (Gilbert, 1989; Quattrone, 1982). Thus, we ex-
pect that accounting information that more transparently conveys
the effects of both outside market forces and managerial action
will allow investors to overcome the tendency to mistakenly blame
or credit managers for ?rm outcomes resulting from exogenous
forces.
Compared to amortized cost-based ?nancial statements, we
propose that reported fair value measurements in the primary
?nancial statements provide information in a more transparent
manner that will reduce the burden of processing and increase
the accuracy of investors’ attributions. First, fair value changes are
timelier than amortized cost changes because they re?ect changes
in condition prior to realization. Second, because fair values re-
?ect exchange prices determined with reference to the aggregated
assumptions of market participants (FASB Accounting Standards
Codi?cation 820-10-05-1B), they are inherently comparable across
?rms. Therefore, observing correlated changes in fair value across
?rms is a direct signal of an exogenous shock to ?rm values. Al-
though exogenous shocks may be systematic and highly correlated
across similar ?rms, shocks arising from managerial actions will
be correlated to a much lesser degree. This differential correlation
enables investors to separate value changes driven by exogenous
forces and managerial actions.
Although amortized cost earnings will also re?ect the real-
ized effects of exogenous shocks to ?rm values, investors must
process earnings innovations to estimate the magnitude and di-
rection of value changes. Because investors’ processing likely in-
troduces estimation error and uncertainty to estimated value
changes, across-?rm correlations will be less transparent, and
manager assessments will be less accurate for investors using
amortized-cost-based ?nancial statements. This leads to our next
hypothesis:
H1b. Fair value information on the face of the ?nancial statements
will mitigate investors’ tendency to positively attribute exogenous
components of ?rm performance to managers.
The interactive effect of recognized and disclosed measurement basis
information on stewardship assessment
We ascribe better manager quality judgments by investors who
receive fair value information to greater transparency (i.e., lower
processing costs) of recognized fair value information than rec-
ognized amortized cost information. The Conceptual Framework
states that in cases where qualitative characteristics are lacking,
“additional disclosures may partially compensate” (FASB, 2010a,
QC34). We therefore also investigate the differential impact of rec-
ognized versus disclosed measurement basis information on stew-
ardship assessment. Speci?cally, we compare fair value recogni-
tion with amortized cost footnotes to amortized cost recognition
with fair value footnotes. We propose that fair value information
provides greater transparency and reduces processing costs to a
greater extent when recognized relative to when it is disclosed
(Maines & McDaniel, 2000; Russo, 1977). Given prior research on
this matter, and supposing fair value recognition transparently
provides information which would allow investors to disentan-
gle market forces from managerial actions, we hypothesize the
following:
H2a. Investors given recognized fair value information on the face
of the ?nancial statements will be more accurate in their manager
quality judgments than investors given fair value information in
the ?nancial statement footnotes.
Because fair value information is more transparent, we propose
that the availability of fair value information in footnotes presents
potential decision-useful bene?ts that should outweigh the cog-
nitive effort necessary to access ?nancial statement footnotes (cf.,
FASB, 2010a; Maines & McDaniel, 2000); thus, we expect improved
performance to the extent investors who view amortized-cost-
based ?nancial statements and who also invest the time and effort
necessary to access fair value information in the footnotes. Stated
more simply, we expect that investors who view amortized-cost-
based information on the face of the ?nancial statements to per-
form better at assessing manager quality to the extent they utilize
the fair value information included in the footnotes. This leads to
the following hypothesis:
104 S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114
H2b. For investors receiving amortized-cost-based ?nancial state-
ments, the accuracy of their manager quality judgments will in-
crease as they acquire and process fair value footnote information.
In contrast to the predicted bene?ts of using fair value foot-
note information, we also anticipate degradation in the accuracy of
investors’ judgments of managers’ quality for investors who view
recognized fair-value-based ?nancial statements to the extent they
access and use amortized cost information disclosed in the foot-
notes. This prediction is consistent with ?ndings in psychology
(e.g., Nisbett et al., 1981) and auditing (Hackenbrack, 1992) that
suggest presenting less diagnostic information causes individuals
to reduce the weight placed on diagnostic information. For this
reason, we predict that the manager quality assessments of in-
vestors viewing fair-value-based ?nancial statements will be worse
to the extent investors utilize amortized cost information in the
footnotes. This leads to the following hypothesis:
H2c. For investors receiving fair-value-based ?nancial statements,
the accuracy of their manager quality judgments will decrease as
they acquire and process amortized-cost-based footnote informa-
tion.
Taken together, H2b and H2c predict an interactive effect on
manager quality based on the extent to which investors access
and use the footnote information. While prior accounting research
has extensively investigated issues related to ?nancial statement
recognition versus footnote disclosure (e.g., Bratten, Choudhary, &
Schipper, 2013), research to date has not investigated how foot-
note disclosure affects the use of recognized ?nancial statement
amounts.
Operationalization of stewardship, experimental setting, design
and procedures
Operationalization of stewardship
As stated earlier, we de?ne stewardship as the actions of man-
agers in the discharge of their responsibilities to preserve the value
of investors’ capital investment and to earn a commensurate return
on that investment. When de?ned in this manner, stewardship can
be considered a reasonable basis on which to reward, punish, or
discharge a business manager.
Because we wish to draw clear inferences about the effect of
measurement basis on investors’ ability to process relevant in-
formation, our operationalization of stewardship is designed to
abstract away from many factors that confound investors’ inter-
pretation of ?nancial performance in naturally occurring settings.
Speci?cally, in our experiment, we create a setting in which in-
vestors know, ex ante, how exogenous shocks and managerial ac-
tions map into ?rm value. Investors also know that an exoge-
nous shock has occurred that was not foreseeable and that man-
agers can only respond to the shock. We operationalize stew-
ardship in terms of managers’ reinvestment decisions because,
as a fundamental managerial duty, these decisions are clearly
relevant to the assessment of managerial performance as well
as asset allocation (Ciesielski, 2013). In the experiment, these
decisions have a direct and proportionate causal effect on fu-
ture realized ?rm earnings, the value of ?rm assets, and ?rm
value. High (low) quality managers reinvest more ?rm earnings in
projects having higher (lower) net present values. Reinvesting in
higher (lower) net present value projects increases (decreases) fu-
ture realized earnings, the fair value of assets, and the value of
the ?rm.
Experimental setting
In our experiment, we examine a setting in which investors
receive ?nancial information containing a balance sheet, income
statement and related footnote disclosures for two ?rms. Both
?rms are in the same industry, and as a result are exposed to the
same industry-wide shocks. Both ?rms have two operating divi-
sions that may be differentially affected by industry-wide shocks.
Investors are informed that the industry is exposed to exogenous
economic shocks that may be either positive or negative with
equal probability, and that such shocks increase or decrease, re-
spectively, the demand for the products of one of the two divi-
sions of each ?rm. Investors are also informed that any shocks
are expected to persist for a period of 5 years.
6
A persistent de-
mand shock is operationalized as a shock to expected future cash
?ows, holding the discount rate constant. This results in a tran-
sient (nonrecurring) change in fair value in the period of the
shock, and an increase or decrease in earned revenue that persists
for 5 years.
Managers respond to shocks by reinvesting prior year’s earnings
into the two divisions. Because the shock is expected to persist into
the future, normatively the relative quality of a manager depends
on the extent to which the manager allocates capital to maximize
?rm value in response to the shock. That is, higher quality man-
agers allocate more earnings to the division with higher future re-
turns than lower quality managers. Each manager’s decision to al-
locate earnings directly affects the ?rm’s value. As such, there are
two factors that affect a ?rm’s value in our setting, the direction of
the exogenous shock and the manager’s capital allocation decision
made in response to the shock.
Design
To test our hypotheses we conduct a computer-based experi-
ment using a 2 d7˜ 3 between-subjects design in which we ma-
nipulate (1) the measurement basis used on the face of the ?nan-
cial statements (full fair value versus amortized cost) and (2) the
assigned judgment objective (investment allocation, stewardship-
quality assessment, or both).
7
In each experimental session, 20 in-
dependent pairs of ?rms in the same industry are presented in
which managers’ reaction (i.e., reinvestment of earnings) to exoge-
nous shocks differs between ?rms. Participants assessing steward-
ship are asked to rate the quality of both ?rm managers, and par-
ticipants making investment allocation decisions are asked to allo-
cate funds between the two ?rms.
Participants include 109 Master of Business Administration and
Master of Accounting students assuming the role of investors.
All participants are randomly assigned to one of the six possi-
ble between-subjects conditions. Participants provide judgments
for one pair of ?rms each round and are compensated based on
the accuracy of their responses. There were 20 rounds in the ex-
periment. After the 20th round, participants complete a brief post-
experimental questionnaire. Amounts in the experiment are ex-
pressed in laboratory currency. At the end of each experimental
6
Because all participants are informed of the expected duration of exogenous
shocks, amortized-cost-based earnings innovations can be transformed with effort
to arrive at equivalent fair value changes (Nissim, 2008). In more complex real-
world settings, unobservable persistence of realized earnings shocks is likely to add
substantial uncertainty to investors’ processing of amortized-cost-based earnings.
Our theory suggests that this would further increase the functional transparency of
fair value measurements relative to amortized cost measurements.
7
We found that making both decisions had no effect on either investors’ stew-
ardship assessments or their asset allocation decisions. At the suggestion of the re-
viewers, we neither distinguish, nor discuss, this additional condition in our re-
ported results. That is, we use these observations in our analysis but do not distin-
guish participants who made both stewardship decisions and investment decisions
from participants who only made one of these decisions.
S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114 105
Fig. 1. Flowchart of study. Notes: Figure presents a ?owchart of the experiment. Participants received a similar ?owchart during the experiment. The ?owchart provided to
the participants only presented the steps that were relevant to their condition. The following box shading/border indicates steps that were assigned only to participants in
the asset allocation, stewardship or both (i.e., joint asset allocation and stewardship) conditions.
session, participants’ cumulative earnings in laboratory currency
are converted to U.S. dollars, and paid to participants in cash at
the end of the session. On average, the experiment took 75 min to
complete, and compensation was approximately $25.00.
Procedures
At the beginning of the study, all participants completed train-
ing on the mechanics of fair value and amortized cost measure-
ment bases in the context of ?xed assets. Participants are then pro-
vided instructions regarding the experiment, including how their
decisions affect their compensation. Next, participants engage in a
practice round, after which they make decisions in 20 independent
rounds.
A summary of each round is presented in Fig. 1. In Step 1, par-
ticipants receive ?nancial statement information and accompany-
ing footnotes for two ?rms for the years 20X1 and 20X2. Both
?rms operate in the same industry and maintain two operating
divisions. The ?nancial statements include either recognized fair
value or amortized cost information, depending on the condition,
with footnotes providing changes in assets and earnings in the op-
posite measurement basis of the one recognized on the ?nancial
statements. For each ?rm, we also present the Chief Executive Of-
?cer’s (CEO’s) name and photograph. Including the CEO name and
photograph allowed participants to personify the managers and is
consistent with the presentation attributes of most published an-
nual reports.
In years 20X1 and 20X2 the earnings across the two divi-
sions are initially automatically equally reinvested into both divi-
sions. However, as illustrated in Step 2 of Fig. 1, an unexpected
exogenous shock occurs at the beginning of 20X3 that is either
positive or negative, and increases or decreases, respectively, the
demand for the products of one of the two divisions of each
?rm. At the end of 20X3 (i.e., after 20X3 results are known by
the manager, but immediately before the 20X3 ?nancial state-
ments are presented to the investors), ?rm managers respond
to this shock by reinvesting 20X3 earnings across ?rm operating
divisions.
Fig. 2 presents an example of a pair of ?rms and accompanying
notes as shown to participants in Step 3. Panel A (Panel B) of Fig. 2
presents the information for one round in an amortized cost (fair
value) condition, where amortized cost (fair value) information is
recognized in the ?nancial statements and fair value (amortized
cost) information is disclosed in the footnotes.
Participants were informed that ?rm managers could not antic-
ipate the shock, and that the magnitude of the shock is expected
to persist for a period of 5 years, thereby affecting the rate of re-
turn for the division’s assets during that period.
8
For example, the
shock to the ?rms in Panels A and B of Fig. 2 is positive. The direc-
tion of the shock is evident in that the sales growth from 20X2 to
20X3 for Division A in each ?rm increased, while the sales growth
for Division B remains unchanged. In addition, in the fair value
condition (Panel B) a revaluation adjustment is made to both Divi-
sion A’s balance sheet and income statement to re?ect the changes
in fair value. Due to market conditions created by the exogenous
shock, the ?rms’ earnings may be greater than or less than the cost
of capital, and investments in productive assets’ earnings above
(below) the cost of capital generate unrealized gains (losses).
9
In addition to providing information regarding the direction and
the division that was affected by the exogenous shock, the ?nan-
cial information for the year 20X3 provides evidence of managers’
8
See Appendix for a detailed discussion regarding the assumptions used in the
valuation of the ?rm.
9
Financial statement information provides inputs for valuation that are helpful
for assessing the amount, timing, and/or risk of future cash ?ows. To achieve the
strongest experimental control, we hold risk and timing of cash ?ows constant and
assess participants’ ability to extract information provided by the ?nancial state-
ments that is relevant for revising expected cash ?ows assessments. We also hold
constant CEO demographics. All CEOs are Caucasian males that appear to be be-
tween the ages of 30 to 60 years old and are randomly assigned within each pair of
?rms. Finally, to avoid order effects, we counterbalance the placement of ?rms on
the left and right side of the screen, and the division to which the shock occurs.
106 S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114
Fig. 2. Example of experimental stimuli.
Fig. 2. Continued
S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114 107
Fig. 2. Continued
allocation of prior year earnings to the two divisions.
10
For exam-
ple, illustrating the greatest difference in manager quality across
?rms, Panels A and B of Fig. 2 show that the manager of Firm A
allocates 100% of prior earnings to the division that did not expe-
10
Although the direction of revenue innovations is unambiguous, the allocation
of depreciation costs in amortized cost accounting may partially obscure the effects
of revenue innovation trends on net income, particularly when nonlinear deprecia-
tion methods are used. Consistent with full fair value recognition, no depreciation
is recognized when fair value information is presented. To minimize the obfuscat-
ing effects of depreciation allocations and any propensity to naively focus on net
income, we (1) apply straight-line depreciation when amortized cost information
is presented and (2) remove net income as a line item in the ?nancial statements.
We believe that both of these design choices bias against the informativeness of
fair value by making more costly any propensity to evaluate performance based
on change in net income. However, we cannot completely eliminate participant-
constructed income ?xation as a contributor to results.
rience a shock (Division B), while the manager for Firm B allocates
100% of earnings to the division that did experience the shock (Di-
vision A). Manager capital allocations are evident in the increase in
gross assets in the respective divisions for 20X3. Because the shock
was positive and thus increased the demand for Division A’s prod-
ucts, the optimal action is to allocate 100% of prior year’s earn-
ings to Division A. As such, the manager for Firm A (Firm B) in
Fig. 2 is classi?ed as lower (higher) in relative stewardship qual-
ity. It is important to note that participants can correctly identify
managers’ relative quality using either amortized cost or fair value
?nancial statements. In both conditions, the change in sales from
20X2 to 20X3 provides an indication of the direction and the divi-
sion that was affected by the shock, and the change in gross assets
from 20X2 to 20X3 allows the identi?cation of the manager’s cap-
ital allocation decision. The primary difference between the two
108 S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114
Fig. 3. Description of ?rm pairs included in the experimental rounds. Note: For each
?rm pair, the “Firm 1” and “Firm 2” were randomly assigned to be on the left side
of the screen or right side of the screen, and were labeled “Firm A” (left side of
screen) or “Firm B” (right side of screen). (See Panel C of Exhibit 2 for an example.)
Additionally, the shock randomly occurred to either Division A or Division B in each
?rm, and the CEO picture and name were randomly assigned to either Firm A or
Firm B for each ?rm pair. There were 42 total names and pictures, so each picture
and name was used exactly once through the one practice round and 20 actual
rounds.
conditions is that the fair value ?nancial statements provide reval-
uation adjustments to income and division assets, which provides
more transparent measures of the impact of the shock as well as
the manager’s capital allocation decision. Fig. 3 provides economic
shock and manager-asset-allocation information for each of the 20
?rm pairs used in the experiment.
11
In Step 3, participants are informed that they lost or earned
compensation based on their investment during the year the shock
occurred. We choose to have participants initially equally invested
(50%) in each ?rm to increase their personal involvement in the
task and to provide a salient incentive to attend to the experimen-
tal task. In Step 4, participants in conditions that include asset al-
location decisions then decide if and how they would like to adjust
the amount of capital allocated between the two ?rms. Participants
can allocate any percentage of their capital between the two ?rms,
and participants’ compensation is based on the percentage of capi-
tal they allocate to the ?rm that would generate greater returns in
the next year.
12
In Step 5, the subset of participants assessing managers’ stew-
ardship makes assessments of each manager’s quality. Panel C of
Fig. 2 provides an example of both the investment and manager
quality decisions participants make. As shown in Panel C of Fig. 2,
participants rate each manager on a scale from “worst possible
quality” to “highest possible quality.” If participants rate one man-
ager greater than another manager, they receive a message con-
?rming that they rated one manager higher than another. Partic-
ipants are also asked how con?dent they are on a scale between
0 and 100 that they have chosen the ?rm with the higher quality
manager. Participants earn (lose) compensation for correctly (incor-
rectly) identifying which manager was of higher quality. The extent
of gain or loss is based on the con?dence they give in their assess-
11
Firm pairs are presented in two different orders. When order is included as a
covariate in the hypotheses tests, it is neither signi?cant nor does it change the
direction or signi?cance of the reported inferential statistics. As such, we do not
include order as a covariate in the reported analyses.
12
The round ends after this step for participants making only asset allocation de-
cisions.
ment.
13
Finally, in Step 6 participants receive feedback in terms of
the compensation they earned during the round based on their de-
cisions.
Results
Descriptive statistics
Table 1 presents the descriptive statistics from our experiment.
Consistent with expectations, participants correctly identi?ed the
higher quality manager in the fair value condition (59%) more of-
ten (p = 0.002) than participants in the amortized cost condi-
tion (53%).
14
We also note that participants in both conditions ac-
cessed the footnote information but participants in the amortized
cost condition spent a higher percentage of time in each round (p
= 0.001) viewing the footnote information (18%) than participants
in the fair value condition (14%).
15
Tests of H1a and H1b
H1a predicts that investors’ assessments of management qual-
ity will be in?uenced by exogenous components of ?rm perfor-
mance. To test H1a, and thereby establish that this attribution
bias exists in our setting, we conduct a regression with MgrQual-
ity as the dependent variable and PosShock as the independent
variable. MgrQuality is the manager rating each participant gave
to a ?rm manager (?100 to 100), with higher values represent-
ing higher perceived manager quality. Within each pair of man-
agers, although one manager is of higher quality than the other,
the extent of difference in quality varies across ?rm pairs, and
there is no normatively correct magnitude for participants’ percep-
tion of quality. Because our aim for H1a and H1b is to test only
whether the perception of manager quality differs as a function of
the exogenous shock, we include PosShock; an indicator variable
that is equal to one (zero) if there was a positive (negative) shock.
A signi?cant coe?cient on PosShock indicates that perceptions of
quality are higher when shocks are positive than when shocks are
negative.
We designed the experiment so that participants would see a
variety of shock–manager-relationships, thereby making it impossi-
ble to heuristically infer systematic patterns in the shock–manager
relationship in the study (e.g., participants could not immediately
assume manager quality was the opposite of the direction of the
shock). However, in our tests of H1a and H1b, we restrict observa-
tions to ?rms that have crossed manager types from shock types.
Speci?cally, we observe cases in which a negative shock occurs to
a ?rm with a high quality manager or a positive shock occurs to a
?rm with a low quality manager. Managers are interpreted as high
(low) quality if they allocated more (less) resources to the ?rm
division with higher returns. The directional difference between
13
The incentives corresponding to participants’ stewardship judgments are pro-
vided as a proxy for an action which investors may take (e.g., a “say on pay” ini-
tiative) which allows voting shareholders to preserve manager contracts and cor-
responding investor returns, or change the principal-agent incentive contract and
allow potential for manager turnover and changes in investment returns.
14
Process data reveal that some participants spent less than 15 seconds in a
round, making it impossible for participants to consider the information contained
on each page. Approximately 90% of these observations were in the last 10 rounds.
We exclude from all analyses a total of 123 (5.6%) out of 2177 observations for
which time spent was less than 15 seconds. Statistical inferences are unchanged
when these observations are included.
15
All regressions are adjusted for repeated measures (clustered at the participant
level). In addition, we conduct all tests using all 20 rounds of data, and separately
using only the last 10 rounds of data to assess the effects of learning on the ad-
vantage of transparency. Statistical inferences do not differ when only the last 10
rounds are evaluated.
S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114 109
Table 1
Descriptive statistics and variable de?nitions.
Amortized cost Fair value
Difference (FV ? AC)
Mann–Whitney t-Test
Cross
MgrQuality, mean, sd 6.91 55.10 13.47 53.81 0.02 0.01
MgrCorrect, mean (se) 52% (0.02) 58% (0.02) 0.02 0.02
InvCorrect, mean (se) 50% (0.02) 54% (0.02) 0.06 0.06
Roundtime, mean, sd 123.84 97.67 113.50 132.09 0.001 0.02
NotesPercent, mean (se) 18% (0.01) 15% (0.01) 0.001 0.001
No Cross
MgrQuality, mean, sd 6.47 54.09 13.14 56.93 0.01 0.01
MgrCorrect, mean (se) 55% (0.02) 60% (0.02) 0.03 0.02
InvCorrect, mean (se) 51% (0.02) 57% (0.02) 0.02 0.02
Roundtime, mean, sd 101.11 81.46 84.05 70.01 0.001 0.001
NotesPercent, mean (se) 18% (0.01) 13% (0.01) 0.001 0.001
All
Participants 55 54
MgrQuality, mean, sd 6.69 54.58 13.31 55.35 0.001 0.001
MgrCorrect, mean (se) 53% (0.01) 59% (0.01) 0.003 0.002
InvCorrect, mean (se) 50% (0.01) 55% (0.01) 0.01 0.004
Roundtime, mean, sd 112.75 90.82 99.05 107.26 0.001 0.001
NotesPercent, mean (se) 18% (0.00) 14% (0.01) 0.001 0.001
MgrQuality, The manager ratings participants provided for each ?rm. Ranges in value from ?100 to 100, with higher values representing higher manager quality. Mgr-
Correct, Indicator variable that re?ects whether participants correctly (one) or incorrectly (zero) identi?ed which manager was higher quality. Roundtime NotesPercent,
The amount of time in seconds spent by participants in each round. The proportion of time spent by participants accessing the notes to the ?nancial statements relative
to the total time spent in each round. InvCorrect, Indicator variable that re?ects whether participants allocated the majority of their funds into the ?rm with higher
(one) or lower (zero) returns. FV, Indicator variable that is equal to one (zero) if investor is presented with Fair Value (Amortized Cost) ?nancial statements. Cross,
Indicator variable that equals one if manager is high (low) quality and exogenous shock is negative (positive), or zero if manager is high (low) quality and exogenous
shock is positive (negative).
shock and manager type allows us to cleanly identify whether par-
ticipants’ quality judgments are in?uenced by the exogenous shock
when they should not be.
16
If the attribution bias we predict exists,
we expect the coe?cient on PosShock to be positive and signi?cant.
Panel A of Table 2 presents our results for our test of H1a. Con-
sistent with H1a, PosShock is positive (5.39) and marginally signif-
icant in the pooled sample (p = 0.098).
17
Manager quality ratings
are more positive when there is a positive shock, which is consis-
tent with attribution bias, and suggests that investors tend to in-
correctly attribute systematic components of ?rm performance to
managers’ actions, even when these components are irrelevant to
actual manager quality. Our theory suggests that systematic bias in
attribution is less likely to occur in the fair value condition, and
likely explains the marginal signi?cance of the pooled results re-
ported in Table 2, Panel A.
H1b predicts that presenting fair values on the face of ?nan-
cial statements will mitigate the tendency to attribute exogenous
components of ?rm performance to managers and will improve
the accuracy of manager quality assessments. To test H1b, we es-
timate the regression used to test H1aand interact PosShock with
FV, where FV is equal to one (zero) if the measurement basis rec-
ognized on the face of the ?nancial statements is fair value (amor-
tized cost). If fair value presentation on the ?nancial statements
16
Because there is no normatively correct magnitude of participants’ quality rat-
ings, a ?nding that participant quality ratings are more positive (negative) for high
quality (low quality) managers when exogenous shocks are positive (negative) could
be consistent either with greater accuracy in assessing manager type, or with the
attribution bias we predict. However, in the case where there is a positive shock
and a low quality manager or a negative shock and a high quality manager (i.e.,
Cross = 1), participants’ manager quality ratings should not be associated with the
direction of the exogenous shock. As such, in our tests of H1a and H1b we focus
on instances where Cross = 1 and interpret an association between subject quality
rating and exogenous shock direction as a bias in attribution.
17
All reported p-values are one-tailed for directional predictions and two-tailed
otherwise.
attenuates attribution errors as predicted by H1b, we expect the
coe?cient on PosShock ? FV to be negative and signi?cant.
Panel B of Table 2 presents results that support H1b. As pre-
dicted, our results reveal that the coe?cient on PosShock ? FV
is negative (?26.18) and signi?cant (p = 0.001). Additionally, the
sum of the coe?cients for PosShock and PosShock ? FV are not
different from zero (untabulated, p = 0.12). Taken together, these
results suggest that recognition of fair value information miti-
gates systematic attribution bias and allows investors to disentan-
gle exogenous and endogenous factors when assessing stewardship
quality.
18
Test of H1b
The results of our ?rst set of hypotheses document the pro-
cessing differences investors use to reach stewardship assess-
ments when given ?nancial statements using differing measure-
ment bases. Speci?cally, we document the existence of an attribu-
tion bias in our setting and provide evidence that fair value recog-
nition ameliorates the bias. Our second set of hypotheses examines
whether the reduction of this bias through fair-value-based recog-
nition leads to more accurate manager quality assessments, and
whether the use of fair value (amortized cost) information in the
notes leads to more (less) accurate manager quality assessments.
For the remainder of our tests, we no longer restrict the dataset to
crossed ?rms, and each ?rm pair constitutes an observation in our
analyses.
18
As an alternate test of H1b, we estimate the regression restricting observations
to Cross = 0, and stack this regression with the tabulated regression in the same
variance–covariance matrix using seemingly unrelated regression techniques. This
speci?cation effectively tests a three-way interaction and allows us to empirically
test whether the difference in manager quality ratings between fair value and amor-
tized cost is driven by cases in which manager and shock types do not match. Unt-
abulated results support H1b (p < 0.01).
110 S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114
Table 2
Tests of H1a and H1b.
Predicted sign Coe?cient Standard error p-Value
Panel A: Regression examining effect of shock on manager quality assessments
MgrQuality = ?
0
+ ?
1
PosShock + b%
Constant 7.57
???
2.60 0.005
PosShock + 5.39
?
4.05 0.094
F-value 1.77
Number of observations 1424
Number of participants 73
Panel B: Effect of shock and measurement basis on manager quality assessments
MgrQuality = ?
0
+ ?
1
PosShock + ?
2
FV + ?
3
PosShock ? FV + b%
Constant ?2.28 3.18 0.476
PosShock 18.65
???
5.85 0.002
FV 19.51
???
4.69 0.001
PosShock ? FV ? ?26.18
???
7.54 0.001
F-value 6.55
Number of observations 1424
Number of participants 73
Notes: Panel A of this table presents the results of a regression model employed
to examine whether the direction of the shock affects participants’ ability to assess
manager quality after controlling for manager quality by restricting observations
to ?rms where manager quality and shock direction differ (i.e., Cross = 1). Panel B
presents the results of a regression model employed to examine whether the mea-
surement basis (either fair value or amortized cost) affects participants’ tendency to
rely on the shock to assess manager quality after restricting observations to ?rms
where manager and shock type differ. Each observation represents an individual
?rm manager. MgrQuality, FV, and PosShock are as de?ned in Table 1. p-Values are
estimated using Huber–White corrected standard errors clustered by participant.
???
Statistical signi?cance at the 1% level.
?
Statistical signi?cance at the 10% level.
To test H2a, we conduct a logistic regression with MgrCorrect
as the dependent variable and FV as the independent variable. The
higher quality manager in each ?rm pair is de?ned as the man-
ager who allocated a larger proportion of prior-period earnings to
the division with a higher net present value. In each ?rm pair, par-
ticipants rate both managers’ relative quality. Using participants’
ratings, our dependent variable MgrCorrect is equal to one (zero)
if a participant correctly (incorrectly) identi?es which ?rm had the
higher quality manager. We expect the coe?cient on FV to be pos-
itive and signi?cant.
Results are presented in Table 3, Panel A. Consistent with H2a,
the coe?cient on FV is positive (0.22) and marginally signi?cant (p
= 0.065).
19
This suggests that fair value recognition more transpar-
ently provides information that allows investors to correctly assess
manager quality relative to amortized cost recognition.
H2b (H2c) predicts that when participants assess stewardship
using ?nancial statements with recognized amortized cost (fair
value) information, the extent to which the participants utilize
supplemental fair value (amortized cost) footnote information will
improve (diminish) their ability to correctly identify higher qual-
ity managers. For H2b and H2c, we conduct a logistic regression
with MgrCorrect as the dependent variable and FV, NotesPercent,
and NotesPercent ? FV as independent variables. NotesPercent is the
percentage of total time that participants spend reviewing the al-
ternative cost basis notes to the ?nancial statements each round.
We expect the interaction coe?cient on NotesPercent ? FV to be
negative and signi?cant.
Table 3 presents the results of our test of H2b and H2c. Consis-
tent with our hypotheses, the coe?cient on NotesPercent ? FV is
negative (?6.80) and signi?cant (p < 0.001), suggesting that as the
19
When restricting our observations to ?rm pairs in which Cross = 1, the positive
coe?cient on FV is signi?cant at conventional levels (p = 0.03).
Fig. 4. Probability of investors identifying the higher quality manager. Note: Fig-
ure displays the probability that participants properly identi?ed the higher qual-
ity manager in our experiment given the percent of time they accessed the foot-
note information. Probabilities are tested for differences in Table 3, Panel B. FV (AC)
refers to conditions in which participants viewed fair value (amortized cost) ?nan-
cial statements with corresponding amortized cost (fair value) information in the
notes.
percentage of time viewing fair value (amortized cost) information
increases, stewardship assessment accuracy increases (decreases).
The signi?cantly positive coe?cient (1.11) on FV (p < 0.001) is
consistent with greater accuracy for participants in the fair value
condition after controlling for notes usage. Untabulated tests for
simple main effects also suggest that both fair value footnote use
and amortized cost footnote use contribute signi?cantly to the in-
teraction (p < 0.01).
To more completely test recognized fair values versus recog-
nized amortized cost with note disclosure of fair values, we com-
pare the probability that the participants identify the higher qual-
ity manager in (1) the case of fair value recognition with no
usage of amortized cost notes (i.e., NotesPercent = 0) versus (2)
the case of amortized cost recognition with the median amount
(10%) of fair value notes usage observed across participants in the
amortized-cost condition. Given these assumed levels of footnote
usage, we ?nd that participants in the fair value condition correctly
identi?ed the higher quality manager (68%) more often (untabu-
lated, p < 0.001) than participants in the amortized cost condition
(50%). These results provide additional support for the assertion
that amortized cost recognition with fair value disclosure is not a
substitute for fair value recognition. We also provide graphical ev-
idence of our interaction in Fig. 4 . As illustrated in Fig. 4, when
investors do not access the footnotes and just use the information
on the face of the ?nancial statements, investors make more ac-
curate stewardship decisions when fair value is recognized com-
pared to when amortized cost is recognized. However, the slope on
the regression line is negative for fair value and positive for amor-
tized cost. Consistent with expectations, the probability of identify-
ing the correct manager decreases as participants spend more time
in the footnotes that contain amortized cost information and in-
creases as participants spend more time in the footnotes that con-
tain fair value information.
Taken together, our tests of H2a–H2c suggest that the recogni-
tion of fair value information leads investors to make better stew-
ardship assessments, and that this result obtains because fair value
information is more transparently diagnostic than amortized cost
information. The actual use of fair value (amortized cost) informa-
tion, whether presented on the face of the ?nancial statements or
disclosed in the notes, leads to better (worse) stewardship assess-
ments. Moreover, disclosure of relevant fair value information is
not a substitute for recognition.
S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114 111
Table 3
Test of H2a, H2b, and H2c.
Predicted sign Relative risk coe?cient Standard error p-Value
Panel A: Effect of measurement basis recognition on participants’ stewardship decisions
P(MgrCorrect) = ?
0
+ ?
1
FV + b%
Constant 0.14
?
0.09 0.055
FV + 0.22
?
0.15 0.065
Wald Chi
2
2.28
Number of observations 1404
Number of participants 73
Panel B: Effect of measurement basis presentation and proportion of time spent accessing notes on participants’ stewardship decisions
P(MgrCorrect) = ?
0
+ ?
1
FV + ?
2
NotesPercent + ?
3
FV ? NotesPercent + b%
Constant ?0.37
???
0.15 0.011
FV + 1.11
???
0.22
Weconductalaboratoryexperimenttoexamineinvestors’assessmentsofmanagers’stewardship.Weprovideevidencethatinvestorstendtoattributeexternal(i.e.,non-manager-related)causesoffirmper-formancetomanagers’performance.Wepredictandfindthatfairvalueinformationenablesinvestorstoovercomethistendencyandmakebetterstewardshipdecisionsthaninvestorswitamortizedcostinformation.Wealsofindthatinvestorspresentedwithamortized-cost-basedfinancialstatementsper-formbettertotheextenttheyaccessfair-value-basedfootnoteinformation,whileinvestorspresentedwithfair-value-basedfinancialstatementsperformworsetotheextenttheyaccessamortized-cost-basedfootnoteinformation.
Accounting, Organizations and Society 46 (2015) 100–114
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
The effect of alternative accounting measurement bases on investors’
assessments of managers’ stewardship
Spencer B. Anderson, Jason L. Brown, Leslie Hodder, Patrick E. Hopkins
?
Kelley School of Business, Indiana University, 1309 E. Tenth Street, Bloomington, IN 47405, United States
a r t i c l e i n f o
Article history:
Received 1 June 2014
Revised 30 March 2015
Accepted 31 March 2015
Available online 29 April 2015
Keywords:
Stewardship
Fair value
Experimental market
Transparency
Comparability
Attribution theory
a b s t r a c t
We conduct a laboratory experiment to examine investors’ assessments of managers’ stewardship. We
provide evidence that investors tend to attribute external (i.e., non-manager-related) causes of ?rm per-
formance to managers’ performance. We predict and ?nd that fair value information enables investors
to overcome this tendency and make better stewardship decisions than investors with amortized cost
information. We also ?nd that investors presented with amortized-cost-based ?nancial statements per-
form better to the extent they access fair-value-based footnote information, while investors presented
with fair-value-based ?nancial statements perform worse to the extent they access amortized-cost-based
footnote information. Collectively, our results suggest that investors’ stewardship decisions are improved
because fair value information more transparently provides the information required to properly consider
the opportunity costs associated with managers’ actions and disentangle endogenous actions by managers
from exogenous market forces that are outside of managers’ control.
© 2015 Elsevier Ltd. All rights reserved.
Introduction
We investigate whether, when holding information constant,
the use of fair value measurements in the primary ?nancial state-
ments improves ?nancial statement users’ (hereafter, “investors’”)
judgments about managers’ stewardship. Opponents of incorporat-
ing fair value measurements into the primary ?nancial statements
argue that fair values do not meet information needs relative to
the stewardship objective of ?nancial reporting, and maintain that
amortized-cost-based ?nancial statements are necessary to assess
stewardship (e.g., Holthausen & Watts, 2001; Nissim & Penman,
2008). We propose that, compared to amortized-cost-based mea-
surement, reported fair value measures in the primary ?nancial
statements provide information in a more transparent manner that
will reduce the burden of processing and improve the accuracy of
stewardship assessments.
As noted by the IASB (2005) staff, the word “stewardship”
has many different meanings in the ?nancial accounting standard-
setting literature. Consistent with the root word “steward,” we pro-
pose that the focus of stewardship is the assessment of actions
taken by stewards (i.e., managers) in the discharge of their respon-
?
Corresponding author.
E-mail addresses: [email protected] (S.B. Anderson), [email protected]
(J.L. Brown), [email protected] (L. Hodder), [email protected] (P.E. Hopkins).
sibilities. In the modern corporation, these responsibilities relate
to an agency relationship and the performance of some delegated
activities on behalf of another party. These delegated activities in-
clude not only safeguarding of assets but also administering busi-
ness resources in a way that generates an acceptable rate of return
on those resources. Therefore, we de?ne stewardship as the actions
of managers in the discharge of their responsibilities to preserve
the value of investors’ capital investment and to earn a commen-
surate return on that investment.
Standard setters currently offer differing positions on the
prominence of stewardship within the overall objective of ?nan-
cial reporting. The FASB recently consolidated the objective of ?-
nancial reporting to resource allocation, subordinating the steward-
ship objective in the process, while stating that information that
is bene?cial for resource allocation is also bene?cial for steward-
ship assessment (FASB, 2010a). Meanwhile, the IASB recently ten-
tatively decided to amend the Conceptual Framework to increase
the prominence of stewardship within the overall objectives of ?-
nancial reporting (IASB, 2014). Our study provides empirical evi-
dence relevant to the stewardship and ?nancial reporting debate,
and addresses O’Connell’s (2007) observation that “stewardship-
related questions have received insu?cient attention in the ?nan-
cial reporting literature to date—particularly from an empirical per-
spective” (p. 216).http://dx.doi.org/10.1016/j.aos.2015.03.007
0361-3682/© 2015 Elsevier Ltd. All rights reserved.
S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114 101
In addition to the controversy over ?nancial reporting objec-
tives, the question of which is the most appropriate measurement
basis for recognized ?nancial statement elements continues to be
one of the most di?cult issues addressed by standard setters. As
either a cause or a consequence of this di?culty, the conceptual
frameworks of the FASB and IASB presently are largely devoid
of any conceptual guidance on accounting measurement issues.
Speci?cally, the FASB Statement of Financial Accounting Concepts
No. 5 (SFAC 5), Recognition and Measurement in Financial Statements
of Business Enterprises (1984, par. 66–70) and the IASB’s Concep-
tual Framework (2014, par. 4.54–4.56) include only terse lists of
measurement bases historically observed in ?nancial statements,
and provide no guidance about the conditions under which one
measurement basis may be preferred over another. According to
Whittington (2008, p. 154), continuing work on the measurement
sections of the conceptual frameworks is likely to be lengthy and
highly contentious because measurement issues often are reduced
to an unproductive competition between fair value and amortized
cost bases of measurement.
1
To correctly assess stewardship, we propose that investors must
(1) receive information that is su?ciently relevant and su?ciently
precise for the task, and (2) disentangle management’s contribu-
tion to ?rm performance from factors outside of management con-
trol (e.g., environmental factors).
2
In this study, we leverage the
comparative advantage of controlled experiments to hold constant
the ?rst requirement to allow us to evaluate the effect of mea-
surement basis on the second. Speci?cally, we conduct a laboratory
experiment to provide empirical evidence relevant to the assess-
ments of stewardship and measurement bases (i.e., amortized cost
and fair value). Our tests are motivated by analytic and psychology
research. Analytic research suggests that high-quality stewardship
assessments depend on principals’ ability to separate the effects of
manager actions from the effects of non-manager (i.e., exogenous)
conditions and events that affect the ?rm (Harris & Raviv, 1978;
Holmström, 1979; Stiglitz, 1974). However, research on attribution
in psychology suggests that investors will have di?culty accurately
assessing managers’ contribution to ?rm performance because in-
dividuals generally have a hard-wired tendency to attribute cause
to individual actors rather than accurately identifying the impact
of exogenous or circumstantial factors (Ross, 1977).
Theory suggests that these tendencies persist because observers
attribute outcomes to salient features of the setting and usually
fail to make the necessary adjustments to correct the automatic
perceptual process (Gilbert, Krull, & Pelham, 1988; Taylor & Fiske,
1975). By making information about exogenous situational factors
more salient, this tendency can be mitigated because less effort
is required to consider situational factors (Gilbert, 1989; Quat-
trone, 1982). We propose that, compared to amortized cost, fair
value measurements provide more transparent information that
can more easily reveal the effects of managerial actions and ex-
ogenous non-manager conditions and events.
Although our primary focus is on measurements recognized in
the primary ?nancial statements, we also investigate the in?u-
ence of notes to the ?nancial statements that include information
about the other measurement bases on investors’ stewardship as-
sessments and investment decisions.
3
Because fair value informa-
1
The debate about the use of fair values in ?nancial statements has been on-
going for most of the 20th Century and has crept into the 21st Century (e.g.,
Chambers, 1966; MacNeal, 1939; Paton & Littleton, 1940; Watts, 2003). As described
by Hodder, Hopkins, and Schipper (2014), many of the arguments on both sides of
the debate remain seemingly intact and untouched by the passage of time.
2
As discussed later in the manuscript, this is consistent with the agency litera-
ture. See Baiman (1982), Hart and Holmström (1987) and Baiman (1990) for com-
prehensive reviews of this literature.
3
Supplemental information about alternative measurement bases is fairly com-
mon in published ?nancial statements (e.g., Statement of Financial Accounting Stan-
tion should be perceived as more transparent, we propose that in-
vestors using amortized-cost-based ?nancial statements will make
higher quality stewardship assessments if they take the time and
effort to process the fair value information in the notes. However,
because amortized cost information provides less salient signals
about exogenous shocks, we expect lower-quality stewardship as-
sessments and investment decisions for users who view fair-value-
based ?nancial statements and spend additional time and effort
processing the amortized-cost-based information in the footnotes
(Hackenbrack, 1992; Nisbett, Zukier, & Lemley, 1981).
In our experiment, participants assume the role of investors
who assess the quality of the ?rm’s management and/or allocate
investment capital to the ?rm. A controlled experiment has ad-
vantages for testing our research questions because we are able
to investigate the effects of a fair value recognition regime that
is largely unobservable in naturally occurring settings and because
we are able to collect information on investors’ decision processes
and judgments that are unavailable in archival databases. Further,
our setting allows us to carefully structure an economic environ-
ment where changes in ?rm value are determined solely by a
combination of (1) exogenous, systematic market forces and (2)
managerial actions conditional on those forces. Ensuring that fun-
damental values and reported results are determined only by a
combination of exogenous shocks and managerial actions provides
greater con?dence that the differences we obtain across measure-
ment bases are strictly a function of investors’ ability to isolate the
contribution of managerial actions to ?rm value.
Participants in the experiment receive information in the form
of ?nancial statements. Approximately half of the participants re-
ceive ?nancial statements in which assets and income are based on
fair value (amortized cost). We also provide to all participants foot-
note information that includes asset and income information using
the measurement basis that is not included on the face of the ?-
nancial statements (i.e., the total information set is the same across
all participants). Speci?cally, participants in the fair-value (amor-
tized cost) conditions can refer to footnote disclosures to obtain
amortized cost (fair value) asset information.
Our results support our prediction that, holding information
constant, recognition of fair values in the primary ?nancial state-
ments enables investors to more accurately assess the quality of
managers’ actions. This is due, at least in part, to the ability of
fair value to ameliorate the tendency to attribute exogenous forces
to management that is apparent when investors view amortized
cost ?nancial statements. Speci?cally, we ?nd that when a nega-
tive (positive) exogenous shock affects ?rm value, investors view-
ing ?nancial statements that report amortized cost measurement
are more likely to believe that ?rm management’s decisions were
of relatively lower (higher) quality, regardless of actual manage-
ment decision quality. For investors viewing fair-value-based ?nan-
cial statements, this bias is mitigated.
By examining process data, we con?rm that better stewardship
assessments and resource allocation decisions are attributable to
investors’ greater (lesser) use of fair value (amortized cost) infor-
mation, even when that information is in the footnotes. Speci?-
cally, investors in the amortized cost condition spending more time
with fair value footnotes performed better relative to investors
spending less time on the fair value footnotes. In contrast, in-
vestors in the fair value condition spending more time on amor-
tized cost footnotes performed worse than investors spending less
time on amortized cost footnotes.
dards No. 107, “Disclosures about Fair Value of Financial Instruments” [FASB 1991]
currently codi?ed in FASB Accounting Standards Codi?cation Topic 825) and in-
creased levels of such disclosure have been included in recent standard-setting pro-
posals (e.g., FASB, 2010b, par. 85).
102 S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114
We organize the rest of this paper as follows: In the next sec-
tion, we outline our theory and hypotheses; in the third section,
we describe our operationalization of stewardship, experimental
task, and research design; in the fourth section we discuss our re-
sults; and in the ?fth section we document our conclusions, de-
scribe important caveats to the generalizability of our study and
provide directions for further research.
Theory and hypotheses
Intersection of accounting measurement and stewardship
As noted by many accounting commentators, a theoretically
neutral starting place for evaluating the relative merits of alterna-
tive measurement bases in the primary ?nancial statements should
begin with identifying the primary objectives of ?nancial reporting.
Following the identi?cation of objectives, the usefulness of alter-
native measurement bases can be assessed relative to these objec-
tives. However, at least some of the literature con?ates these two
steps by holding self-evident the joint propositions that (1) stew-
ardship is the most important objective of ?nancial reporting and
(2) good stewardship can be achieved only with amortized-cost-
based ?nancial statements (e.g., Birnberg, 1980; Chen, 1975).
Why stewardship and amortized cost are so closely linked
is unclear; however, several authors assert that amortized cost,
transaction-based accounting information is desirable primarily
because it is more veri?able (e.g., Demski & Sappington, 1993;
Watts, 2003). Researchers have noted the linkages between stew-
ardship and agency theory (Baiman, 1990; Birnberg, 1980; Gjes-
dal, 1981). More recent research suggests that information rel-
evant for incentive and control purposes necessarily focuses on
completed transactions and past performance, rather than esti-
mates and future-oriented projections (Ball, 2001; Barclay, Gode,
& Kothari, 2005). These arguments shift the focus of the measure-
ment basis debate by introducing a more speci?c set of poten-
tially testable, but largely untested, assertions: (1) that amortized-
cost-based accounting information is inherently more veri?-
able than fair-value-based accounting information and (2) that
amortized-cost-based ?nancial information is inherently more con-
tractible than fair-value-based ?nancial information in an agency
setting.
There are reasons to question—and for scholars ultimately to
test—both of these assertions. First, despite persistent claims (e.g.,
Ramanna & Watts, 2012), research provides no empirical evidence
that fair values are inherently less veri?able than amortized costs
(see discussion in Hodder et al. (2014)). Although fair valuation po-
tentially introduces measurement uncertainty into ?nancial state-
ments, the practice of deferring and amortizing costs potentially
introduces both measurement uncertainty and existence uncer-
tainty. This is because amortized cost ?nancial statements are
based on a set of accruals, deferrals, and allocations, the objective
appropriateness of which is di?cult to independently verify.
4
As
Glover, Ijiri, Levine, and Liang (2005) acknowledge, ?nancial infor-
mation can be classi?ed along a continuum of veri?ability across
all measurement bases. Along this continuum, some fair-value-
based measurements are likely to be more veri?able than some
amortized-cost-based measurements. For example, the change in
fair value of a marketable ?nancial instrument may be much eas-
ier to verify than whether revenue has been earned in a particu-
lar period. Gjesdal (1981) recognizes that veri?ability is not abso-
lute, and establishes that information useful for stewardship must
4
For example, although an expenditure of cash may be easily veri?ed, there may
be substantial uncertainty about the existence of future economic bene?ts that may
be created by that expenditure, or whether the value of those bene?ts, if they exist,
is equal to the accumulated or allocated costs expended.
be only su?ciently veri?able at a reasonable cost. Because both
fair values and amortized costs may meet these criteria, veri?abil-
ity cannot be a su?cient justi?cation for rejecting fair value as a
measurement basis useful for stewardship.
Second, although evaluating managers based on realized ob-
servable outcomes facilitates “settling up” that may prevent ex post
inequitable incentive payments (Leone, Wu, & Zimmerman, 2006),
such contracts may not incent optimal ex ante actions, and these
ex ante actions more generally are the focus of contracting in an
agency setting. For example, if incentive compensation is based on
realizations, and outcome risk is su?ciently high, then risk-averse
managers may be unwilling to accept projects that clearly are in
the best interests of the shareholders (Weiss, 2011). In contrast
to the view that performance information should facilitate observ-
able outcome-based settling up, the agency literature establishes
that performance information is most useful for control when it
reduces uncertainty about agents’ current actions, not necessarily
current results (Gjesdal, 1981; Paul, 1992). If changes in fair val-
ues re?ect the present value of managerial actions, then fair values
may be useful for contracting precisely because they are not real-
ized outcomes, and instead re?ect forward-looking aggregations of
currently available information. If contracts based on “settling up”
are not necessarily more e?cient than contracts based on aggre-
gations of currently available information, the fact that fair values
re?ect forward-looking information cannot be a su?cient justi?-
cation for rejecting fair value measurement as a basis useful for
stewardship.
The fact that veri?ability and ex post settling-up value are
not su?cient reasons for rejecting fair value as a stewardship-
relevant measurement basis also does not imply that amortized
cost-basis measurements should be rejected in favor of fair value
measurements. We propose that identifying the conditions under
which fair-value-based or amortized-cost-based measurements are
more useful for stewardship is an empirical question. According
to agency theory, the answer to this question will be a function
of which measurement basis best helps investors isolate the ef-
fects of managerial actions on ?rm value. As Paul (1992) sum-
marizes, “to provide optimal incentives in a principal-agent prob-
lem, we need to weight information according to its informative-
ness about the manager’s contribution, or value-added, to the ?rm”
(p. 472).
Both amortized cost and fair value measurement bases produce
summary outputs that re?ect the effects of managerial actions;
however, both measures also confound the effects of managerial
actions with other exogenous events that impact the ?rm. For
this reason, we posit that, holding information constant, the rel-
ative usefulness of alternative measurement bases in the primary
?nancial statements must be a function of transparency. We de-
?ne transparency loosely as ?nancial statement presentation that
facilitates the extraction of decision-relevant information.
5
Speci?-
cally, even if the contribution of the manager theoretically can be
extracted from aggregated ?nancial results by appropriate use of
all recognized and disclosed information, investors may improp-
erly attribute variation in ?rm performance to variation in man-
ager actions if the effect of those actions is not su?ciently trans-
parent. This is consistent with psychology research on attribution,
which ?nds that individuals have a natural tendency to attribute
cause to individuals, rather than considering exogenous or circum-
stantial factors (Ross, 1977). We discuss this theory in the next
section.
5
Fair value measurement also has other attributes that may increase informative-
ness relative to amortized costs. For example, the greater timeliness of fair values
is well documented (see Laux and Leuz (2009) for a discussion).
S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114 103
Attribution theory
Incorrectly attributing changes in company performance to
manager actions undermines the e?ciency of incentive contracts
that are based on ?nancial results (Gjesdal, 1981; Paul, 1992). Sev-
eral models exist to describe the attribution process, but all gen-
erally agree that the process of attributing cause to events can be
broken into two steps: ?rst, attributions are initially automatically
activated based on observed stimuli, and second, more controlled
processes may override initial automatic judgments when activated
(Gilbert et al., 1988; Quattrone, 1982; Trope, 1986; Winter & Ule-
man, 1984). These models also posit that initial attributional judg-
ments on individuals can be incorrect, particularly when multiple
causes for behavior are at play and ambiguity shrouds the cause–
effect relationship.
Extant research on causal inference in psychology suggests that
when observers initially attribute causes for outcomes, they tend
to overweight personal dispositions in human actors (Ross, 1977).
That is, in an environment where an outcome is dependent on
two factors, one based on manager actions and the other resulting
from an exogenous in?uence, individuals will overweight manage-
ment’s role and underweight the role of exogenous factors. Consis-
tent with this theory, we predict that when a report of ?rm per-
formance is provided, investors will be prone to attribute cause to
managers, irrespective of managers’ contribution. This leads to our
?rst hypothesis:
H1a. In their manager quality assessments, investors will posi-
tively attribute exogenous (i.e., non-manager) components of ?rm
performance to managers.
Effect of measurement bases on attribution errors
To overcome the tendency of investors to incorrectly attribute
exogenous components of ?rm performance to managers’ actions,
investors must correctly identify the direction and magnitude of
managerial actions on ?rm performance, independent of the im-
pact of exogenous factors. However, because the tendency to at-
tribute cause for ?rm outcomes to managers is automatic, over-
coming this potential bias may be di?cult (Winter & Uleman,
1984). These tendencies that result in attribution errors are theo-
rized to persist because (1) observers attribute outcomes to poten-
tial causes that capture attention and (2) individuals fail to make a
deliberate and conscious effort to adjust original automatic causal
attributions (Gilbert et al., 1988; Taylor & Fiske, 1975).
One way to correct for investors’ tendency to attribute causes
of ?rm outcomes to managers is by increasing the transparency of
exogenous (i.e., non-manager-related) factors. In most settings, in-
dividuals’ attributions focus on individuals rather than situational
factors that are more di?cult to visualize (Lassiter, Geers, Munhall,
Ploutz-Snyder, & Breitenbecher, 2002). Therefore, when informa-
tion regarding situational factors at play is provided more transpar-
ently, systematic attribution errors can be corrected by (1) shift-
ing the perceptual focus of potential causes to exogenous forces
(Gilbert, 2002) and (2) decreasing the effort required to consider
situational factors (Gilbert, 1989; Quattrone, 1982). Thus, we ex-
pect that accounting information that more transparently conveys
the effects of both outside market forces and managerial action
will allow investors to overcome the tendency to mistakenly blame
or credit managers for ?rm outcomes resulting from exogenous
forces.
Compared to amortized cost-based ?nancial statements, we
propose that reported fair value measurements in the primary
?nancial statements provide information in a more transparent
manner that will reduce the burden of processing and increase
the accuracy of investors’ attributions. First, fair value changes are
timelier than amortized cost changes because they re?ect changes
in condition prior to realization. Second, because fair values re-
?ect exchange prices determined with reference to the aggregated
assumptions of market participants (FASB Accounting Standards
Codi?cation 820-10-05-1B), they are inherently comparable across
?rms. Therefore, observing correlated changes in fair value across
?rms is a direct signal of an exogenous shock to ?rm values. Al-
though exogenous shocks may be systematic and highly correlated
across similar ?rms, shocks arising from managerial actions will
be correlated to a much lesser degree. This differential correlation
enables investors to separate value changes driven by exogenous
forces and managerial actions.
Although amortized cost earnings will also re?ect the real-
ized effects of exogenous shocks to ?rm values, investors must
process earnings innovations to estimate the magnitude and di-
rection of value changes. Because investors’ processing likely in-
troduces estimation error and uncertainty to estimated value
changes, across-?rm correlations will be less transparent, and
manager assessments will be less accurate for investors using
amortized-cost-based ?nancial statements. This leads to our next
hypothesis:
H1b. Fair value information on the face of the ?nancial statements
will mitigate investors’ tendency to positively attribute exogenous
components of ?rm performance to managers.
The interactive effect of recognized and disclosed measurement basis
information on stewardship assessment
We ascribe better manager quality judgments by investors who
receive fair value information to greater transparency (i.e., lower
processing costs) of recognized fair value information than rec-
ognized amortized cost information. The Conceptual Framework
states that in cases where qualitative characteristics are lacking,
“additional disclosures may partially compensate” (FASB, 2010a,
QC34). We therefore also investigate the differential impact of rec-
ognized versus disclosed measurement basis information on stew-
ardship assessment. Speci?cally, we compare fair value recogni-
tion with amortized cost footnotes to amortized cost recognition
with fair value footnotes. We propose that fair value information
provides greater transparency and reduces processing costs to a
greater extent when recognized relative to when it is disclosed
(Maines & McDaniel, 2000; Russo, 1977). Given prior research on
this matter, and supposing fair value recognition transparently
provides information which would allow investors to disentan-
gle market forces from managerial actions, we hypothesize the
following:
H2a. Investors given recognized fair value information on the face
of the ?nancial statements will be more accurate in their manager
quality judgments than investors given fair value information in
the ?nancial statement footnotes.
Because fair value information is more transparent, we propose
that the availability of fair value information in footnotes presents
potential decision-useful bene?ts that should outweigh the cog-
nitive effort necessary to access ?nancial statement footnotes (cf.,
FASB, 2010a; Maines & McDaniel, 2000); thus, we expect improved
performance to the extent investors who view amortized-cost-
based ?nancial statements and who also invest the time and effort
necessary to access fair value information in the footnotes. Stated
more simply, we expect that investors who view amortized-cost-
based information on the face of the ?nancial statements to per-
form better at assessing manager quality to the extent they utilize
the fair value information included in the footnotes. This leads to
the following hypothesis:
104 S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114
H2b. For investors receiving amortized-cost-based ?nancial state-
ments, the accuracy of their manager quality judgments will in-
crease as they acquire and process fair value footnote information.
In contrast to the predicted bene?ts of using fair value foot-
note information, we also anticipate degradation in the accuracy of
investors’ judgments of managers’ quality for investors who view
recognized fair-value-based ?nancial statements to the extent they
access and use amortized cost information disclosed in the foot-
notes. This prediction is consistent with ?ndings in psychology
(e.g., Nisbett et al., 1981) and auditing (Hackenbrack, 1992) that
suggest presenting less diagnostic information causes individuals
to reduce the weight placed on diagnostic information. For this
reason, we predict that the manager quality assessments of in-
vestors viewing fair-value-based ?nancial statements will be worse
to the extent investors utilize amortized cost information in the
footnotes. This leads to the following hypothesis:
H2c. For investors receiving fair-value-based ?nancial statements,
the accuracy of their manager quality judgments will decrease as
they acquire and process amortized-cost-based footnote informa-
tion.
Taken together, H2b and H2c predict an interactive effect on
manager quality based on the extent to which investors access
and use the footnote information. While prior accounting research
has extensively investigated issues related to ?nancial statement
recognition versus footnote disclosure (e.g., Bratten, Choudhary, &
Schipper, 2013), research to date has not investigated how foot-
note disclosure affects the use of recognized ?nancial statement
amounts.
Operationalization of stewardship, experimental setting, design
and procedures
Operationalization of stewardship
As stated earlier, we de?ne stewardship as the actions of man-
agers in the discharge of their responsibilities to preserve the value
of investors’ capital investment and to earn a commensurate return
on that investment. When de?ned in this manner, stewardship can
be considered a reasonable basis on which to reward, punish, or
discharge a business manager.
Because we wish to draw clear inferences about the effect of
measurement basis on investors’ ability to process relevant in-
formation, our operationalization of stewardship is designed to
abstract away from many factors that confound investors’ inter-
pretation of ?nancial performance in naturally occurring settings.
Speci?cally, in our experiment, we create a setting in which in-
vestors know, ex ante, how exogenous shocks and managerial ac-
tions map into ?rm value. Investors also know that an exoge-
nous shock has occurred that was not foreseeable and that man-
agers can only respond to the shock. We operationalize stew-
ardship in terms of managers’ reinvestment decisions because,
as a fundamental managerial duty, these decisions are clearly
relevant to the assessment of managerial performance as well
as asset allocation (Ciesielski, 2013). In the experiment, these
decisions have a direct and proportionate causal effect on fu-
ture realized ?rm earnings, the value of ?rm assets, and ?rm
value. High (low) quality managers reinvest more ?rm earnings in
projects having higher (lower) net present values. Reinvesting in
higher (lower) net present value projects increases (decreases) fu-
ture realized earnings, the fair value of assets, and the value of
the ?rm.
Experimental setting
In our experiment, we examine a setting in which investors
receive ?nancial information containing a balance sheet, income
statement and related footnote disclosures for two ?rms. Both
?rms are in the same industry, and as a result are exposed to the
same industry-wide shocks. Both ?rms have two operating divi-
sions that may be differentially affected by industry-wide shocks.
Investors are informed that the industry is exposed to exogenous
economic shocks that may be either positive or negative with
equal probability, and that such shocks increase or decrease, re-
spectively, the demand for the products of one of the two divi-
sions of each ?rm. Investors are also informed that any shocks
are expected to persist for a period of 5 years.
6
A persistent de-
mand shock is operationalized as a shock to expected future cash
?ows, holding the discount rate constant. This results in a tran-
sient (nonrecurring) change in fair value in the period of the
shock, and an increase or decrease in earned revenue that persists
for 5 years.
Managers respond to shocks by reinvesting prior year’s earnings
into the two divisions. Because the shock is expected to persist into
the future, normatively the relative quality of a manager depends
on the extent to which the manager allocates capital to maximize
?rm value in response to the shock. That is, higher quality man-
agers allocate more earnings to the division with higher future re-
turns than lower quality managers. Each manager’s decision to al-
locate earnings directly affects the ?rm’s value. As such, there are
two factors that affect a ?rm’s value in our setting, the direction of
the exogenous shock and the manager’s capital allocation decision
made in response to the shock.
Design
To test our hypotheses we conduct a computer-based experi-
ment using a 2 d7˜ 3 between-subjects design in which we ma-
nipulate (1) the measurement basis used on the face of the ?nan-
cial statements (full fair value versus amortized cost) and (2) the
assigned judgment objective (investment allocation, stewardship-
quality assessment, or both).
7
In each experimental session, 20 in-
dependent pairs of ?rms in the same industry are presented in
which managers’ reaction (i.e., reinvestment of earnings) to exoge-
nous shocks differs between ?rms. Participants assessing steward-
ship are asked to rate the quality of both ?rm managers, and par-
ticipants making investment allocation decisions are asked to allo-
cate funds between the two ?rms.
Participants include 109 Master of Business Administration and
Master of Accounting students assuming the role of investors.
All participants are randomly assigned to one of the six possi-
ble between-subjects conditions. Participants provide judgments
for one pair of ?rms each round and are compensated based on
the accuracy of their responses. There were 20 rounds in the ex-
periment. After the 20th round, participants complete a brief post-
experimental questionnaire. Amounts in the experiment are ex-
pressed in laboratory currency. At the end of each experimental
6
Because all participants are informed of the expected duration of exogenous
shocks, amortized-cost-based earnings innovations can be transformed with effort
to arrive at equivalent fair value changes (Nissim, 2008). In more complex real-
world settings, unobservable persistence of realized earnings shocks is likely to add
substantial uncertainty to investors’ processing of amortized-cost-based earnings.
Our theory suggests that this would further increase the functional transparency of
fair value measurements relative to amortized cost measurements.
7
We found that making both decisions had no effect on either investors’ stew-
ardship assessments or their asset allocation decisions. At the suggestion of the re-
viewers, we neither distinguish, nor discuss, this additional condition in our re-
ported results. That is, we use these observations in our analysis but do not distin-
guish participants who made both stewardship decisions and investment decisions
from participants who only made one of these decisions.
S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114 105
Fig. 1. Flowchart of study. Notes: Figure presents a ?owchart of the experiment. Participants received a similar ?owchart during the experiment. The ?owchart provided to
the participants only presented the steps that were relevant to their condition. The following box shading/border indicates steps that were assigned only to participants in
the asset allocation, stewardship or both (i.e., joint asset allocation and stewardship) conditions.
session, participants’ cumulative earnings in laboratory currency
are converted to U.S. dollars, and paid to participants in cash at
the end of the session. On average, the experiment took 75 min to
complete, and compensation was approximately $25.00.
Procedures
At the beginning of the study, all participants completed train-
ing on the mechanics of fair value and amortized cost measure-
ment bases in the context of ?xed assets. Participants are then pro-
vided instructions regarding the experiment, including how their
decisions affect their compensation. Next, participants engage in a
practice round, after which they make decisions in 20 independent
rounds.
A summary of each round is presented in Fig. 1. In Step 1, par-
ticipants receive ?nancial statement information and accompany-
ing footnotes for two ?rms for the years 20X1 and 20X2. Both
?rms operate in the same industry and maintain two operating
divisions. The ?nancial statements include either recognized fair
value or amortized cost information, depending on the condition,
with footnotes providing changes in assets and earnings in the op-
posite measurement basis of the one recognized on the ?nancial
statements. For each ?rm, we also present the Chief Executive Of-
?cer’s (CEO’s) name and photograph. Including the CEO name and
photograph allowed participants to personify the managers and is
consistent with the presentation attributes of most published an-
nual reports.
In years 20X1 and 20X2 the earnings across the two divi-
sions are initially automatically equally reinvested into both divi-
sions. However, as illustrated in Step 2 of Fig. 1, an unexpected
exogenous shock occurs at the beginning of 20X3 that is either
positive or negative, and increases or decreases, respectively, the
demand for the products of one of the two divisions of each
?rm. At the end of 20X3 (i.e., after 20X3 results are known by
the manager, but immediately before the 20X3 ?nancial state-
ments are presented to the investors), ?rm managers respond
to this shock by reinvesting 20X3 earnings across ?rm operating
divisions.
Fig. 2 presents an example of a pair of ?rms and accompanying
notes as shown to participants in Step 3. Panel A (Panel B) of Fig. 2
presents the information for one round in an amortized cost (fair
value) condition, where amortized cost (fair value) information is
recognized in the ?nancial statements and fair value (amortized
cost) information is disclosed in the footnotes.
Participants were informed that ?rm managers could not antic-
ipate the shock, and that the magnitude of the shock is expected
to persist for a period of 5 years, thereby affecting the rate of re-
turn for the division’s assets during that period.
8
For example, the
shock to the ?rms in Panels A and B of Fig. 2 is positive. The direc-
tion of the shock is evident in that the sales growth from 20X2 to
20X3 for Division A in each ?rm increased, while the sales growth
for Division B remains unchanged. In addition, in the fair value
condition (Panel B) a revaluation adjustment is made to both Divi-
sion A’s balance sheet and income statement to re?ect the changes
in fair value. Due to market conditions created by the exogenous
shock, the ?rms’ earnings may be greater than or less than the cost
of capital, and investments in productive assets’ earnings above
(below) the cost of capital generate unrealized gains (losses).
9
In addition to providing information regarding the direction and
the division that was affected by the exogenous shock, the ?nan-
cial information for the year 20X3 provides evidence of managers’
8
See Appendix for a detailed discussion regarding the assumptions used in the
valuation of the ?rm.
9
Financial statement information provides inputs for valuation that are helpful
for assessing the amount, timing, and/or risk of future cash ?ows. To achieve the
strongest experimental control, we hold risk and timing of cash ?ows constant and
assess participants’ ability to extract information provided by the ?nancial state-
ments that is relevant for revising expected cash ?ows assessments. We also hold
constant CEO demographics. All CEOs are Caucasian males that appear to be be-
tween the ages of 30 to 60 years old and are randomly assigned within each pair of
?rms. Finally, to avoid order effects, we counterbalance the placement of ?rms on
the left and right side of the screen, and the division to which the shock occurs.
106 S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114
Fig. 2. Example of experimental stimuli.
Fig. 2. Continued
S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114 107
Fig. 2. Continued
allocation of prior year earnings to the two divisions.
10
For exam-
ple, illustrating the greatest difference in manager quality across
?rms, Panels A and B of Fig. 2 show that the manager of Firm A
allocates 100% of prior earnings to the division that did not expe-
10
Although the direction of revenue innovations is unambiguous, the allocation
of depreciation costs in amortized cost accounting may partially obscure the effects
of revenue innovation trends on net income, particularly when nonlinear deprecia-
tion methods are used. Consistent with full fair value recognition, no depreciation
is recognized when fair value information is presented. To minimize the obfuscat-
ing effects of depreciation allocations and any propensity to naively focus on net
income, we (1) apply straight-line depreciation when amortized cost information
is presented and (2) remove net income as a line item in the ?nancial statements.
We believe that both of these design choices bias against the informativeness of
fair value by making more costly any propensity to evaluate performance based
on change in net income. However, we cannot completely eliminate participant-
constructed income ?xation as a contributor to results.
rience a shock (Division B), while the manager for Firm B allocates
100% of earnings to the division that did experience the shock (Di-
vision A). Manager capital allocations are evident in the increase in
gross assets in the respective divisions for 20X3. Because the shock
was positive and thus increased the demand for Division A’s prod-
ucts, the optimal action is to allocate 100% of prior year’s earn-
ings to Division A. As such, the manager for Firm A (Firm B) in
Fig. 2 is classi?ed as lower (higher) in relative stewardship qual-
ity. It is important to note that participants can correctly identify
managers’ relative quality using either amortized cost or fair value
?nancial statements. In both conditions, the change in sales from
20X2 to 20X3 provides an indication of the direction and the divi-
sion that was affected by the shock, and the change in gross assets
from 20X2 to 20X3 allows the identi?cation of the manager’s cap-
ital allocation decision. The primary difference between the two
108 S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114
Fig. 3. Description of ?rm pairs included in the experimental rounds. Note: For each
?rm pair, the “Firm 1” and “Firm 2” were randomly assigned to be on the left side
of the screen or right side of the screen, and were labeled “Firm A” (left side of
screen) or “Firm B” (right side of screen). (See Panel C of Exhibit 2 for an example.)
Additionally, the shock randomly occurred to either Division A or Division B in each
?rm, and the CEO picture and name were randomly assigned to either Firm A or
Firm B for each ?rm pair. There were 42 total names and pictures, so each picture
and name was used exactly once through the one practice round and 20 actual
rounds.
conditions is that the fair value ?nancial statements provide reval-
uation adjustments to income and division assets, which provides
more transparent measures of the impact of the shock as well as
the manager’s capital allocation decision. Fig. 3 provides economic
shock and manager-asset-allocation information for each of the 20
?rm pairs used in the experiment.
11
In Step 3, participants are informed that they lost or earned
compensation based on their investment during the year the shock
occurred. We choose to have participants initially equally invested
(50%) in each ?rm to increase their personal involvement in the
task and to provide a salient incentive to attend to the experimen-
tal task. In Step 4, participants in conditions that include asset al-
location decisions then decide if and how they would like to adjust
the amount of capital allocated between the two ?rms. Participants
can allocate any percentage of their capital between the two ?rms,
and participants’ compensation is based on the percentage of capi-
tal they allocate to the ?rm that would generate greater returns in
the next year.
12
In Step 5, the subset of participants assessing managers’ stew-
ardship makes assessments of each manager’s quality. Panel C of
Fig. 2 provides an example of both the investment and manager
quality decisions participants make. As shown in Panel C of Fig. 2,
participants rate each manager on a scale from “worst possible
quality” to “highest possible quality.” If participants rate one man-
ager greater than another manager, they receive a message con-
?rming that they rated one manager higher than another. Partic-
ipants are also asked how con?dent they are on a scale between
0 and 100 that they have chosen the ?rm with the higher quality
manager. Participants earn (lose) compensation for correctly (incor-
rectly) identifying which manager was of higher quality. The extent
of gain or loss is based on the con?dence they give in their assess-
11
Firm pairs are presented in two different orders. When order is included as a
covariate in the hypotheses tests, it is neither signi?cant nor does it change the
direction or signi?cance of the reported inferential statistics. As such, we do not
include order as a covariate in the reported analyses.
12
The round ends after this step for participants making only asset allocation de-
cisions.
ment.
13
Finally, in Step 6 participants receive feedback in terms of
the compensation they earned during the round based on their de-
cisions.
Results
Descriptive statistics
Table 1 presents the descriptive statistics from our experiment.
Consistent with expectations, participants correctly identi?ed the
higher quality manager in the fair value condition (59%) more of-
ten (p = 0.002) than participants in the amortized cost condi-
tion (53%).
14
We also note that participants in both conditions ac-
cessed the footnote information but participants in the amortized
cost condition spent a higher percentage of time in each round (p
= 0.001) viewing the footnote information (18%) than participants
in the fair value condition (14%).
15
Tests of H1a and H1b
H1a predicts that investors’ assessments of management qual-
ity will be in?uenced by exogenous components of ?rm perfor-
mance. To test H1a, and thereby establish that this attribution
bias exists in our setting, we conduct a regression with MgrQual-
ity as the dependent variable and PosShock as the independent
variable. MgrQuality is the manager rating each participant gave
to a ?rm manager (?100 to 100), with higher values represent-
ing higher perceived manager quality. Within each pair of man-
agers, although one manager is of higher quality than the other,
the extent of difference in quality varies across ?rm pairs, and
there is no normatively correct magnitude for participants’ percep-
tion of quality. Because our aim for H1a and H1b is to test only
whether the perception of manager quality differs as a function of
the exogenous shock, we include PosShock; an indicator variable
that is equal to one (zero) if there was a positive (negative) shock.
A signi?cant coe?cient on PosShock indicates that perceptions of
quality are higher when shocks are positive than when shocks are
negative.
We designed the experiment so that participants would see a
variety of shock–manager-relationships, thereby making it impossi-
ble to heuristically infer systematic patterns in the shock–manager
relationship in the study (e.g., participants could not immediately
assume manager quality was the opposite of the direction of the
shock). However, in our tests of H1a and H1b, we restrict observa-
tions to ?rms that have crossed manager types from shock types.
Speci?cally, we observe cases in which a negative shock occurs to
a ?rm with a high quality manager or a positive shock occurs to a
?rm with a low quality manager. Managers are interpreted as high
(low) quality if they allocated more (less) resources to the ?rm
division with higher returns. The directional difference between
13
The incentives corresponding to participants’ stewardship judgments are pro-
vided as a proxy for an action which investors may take (e.g., a “say on pay” ini-
tiative) which allows voting shareholders to preserve manager contracts and cor-
responding investor returns, or change the principal-agent incentive contract and
allow potential for manager turnover and changes in investment returns.
14
Process data reveal that some participants spent less than 15 seconds in a
round, making it impossible for participants to consider the information contained
on each page. Approximately 90% of these observations were in the last 10 rounds.
We exclude from all analyses a total of 123 (5.6%) out of 2177 observations for
which time spent was less than 15 seconds. Statistical inferences are unchanged
when these observations are included.
15
All regressions are adjusted for repeated measures (clustered at the participant
level). In addition, we conduct all tests using all 20 rounds of data, and separately
using only the last 10 rounds of data to assess the effects of learning on the ad-
vantage of transparency. Statistical inferences do not differ when only the last 10
rounds are evaluated.
S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114 109
Table 1
Descriptive statistics and variable de?nitions.
Amortized cost Fair value
Difference (FV ? AC)
Mann–Whitney t-Test
Cross
MgrQuality, mean, sd 6.91 55.10 13.47 53.81 0.02 0.01
MgrCorrect, mean (se) 52% (0.02) 58% (0.02) 0.02 0.02
InvCorrect, mean (se) 50% (0.02) 54% (0.02) 0.06 0.06
Roundtime, mean, sd 123.84 97.67 113.50 132.09 0.001 0.02
NotesPercent, mean (se) 18% (0.01) 15% (0.01) 0.001 0.001
No Cross
MgrQuality, mean, sd 6.47 54.09 13.14 56.93 0.01 0.01
MgrCorrect, mean (se) 55% (0.02) 60% (0.02) 0.03 0.02
InvCorrect, mean (se) 51% (0.02) 57% (0.02) 0.02 0.02
Roundtime, mean, sd 101.11 81.46 84.05 70.01 0.001 0.001
NotesPercent, mean (se) 18% (0.01) 13% (0.01) 0.001 0.001
All
Participants 55 54
MgrQuality, mean, sd 6.69 54.58 13.31 55.35 0.001 0.001
MgrCorrect, mean (se) 53% (0.01) 59% (0.01) 0.003 0.002
InvCorrect, mean (se) 50% (0.01) 55% (0.01) 0.01 0.004
Roundtime, mean, sd 112.75 90.82 99.05 107.26 0.001 0.001
NotesPercent, mean (se) 18% (0.00) 14% (0.01) 0.001 0.001
MgrQuality, The manager ratings participants provided for each ?rm. Ranges in value from ?100 to 100, with higher values representing higher manager quality. Mgr-
Correct, Indicator variable that re?ects whether participants correctly (one) or incorrectly (zero) identi?ed which manager was higher quality. Roundtime NotesPercent,
The amount of time in seconds spent by participants in each round. The proportion of time spent by participants accessing the notes to the ?nancial statements relative
to the total time spent in each round. InvCorrect, Indicator variable that re?ects whether participants allocated the majority of their funds into the ?rm with higher
(one) or lower (zero) returns. FV, Indicator variable that is equal to one (zero) if investor is presented with Fair Value (Amortized Cost) ?nancial statements. Cross,
Indicator variable that equals one if manager is high (low) quality and exogenous shock is negative (positive), or zero if manager is high (low) quality and exogenous
shock is positive (negative).
shock and manager type allows us to cleanly identify whether par-
ticipants’ quality judgments are in?uenced by the exogenous shock
when they should not be.
16
If the attribution bias we predict exists,
we expect the coe?cient on PosShock to be positive and signi?cant.
Panel A of Table 2 presents our results for our test of H1a. Con-
sistent with H1a, PosShock is positive (5.39) and marginally signif-
icant in the pooled sample (p = 0.098).
17
Manager quality ratings
are more positive when there is a positive shock, which is consis-
tent with attribution bias, and suggests that investors tend to in-
correctly attribute systematic components of ?rm performance to
managers’ actions, even when these components are irrelevant to
actual manager quality. Our theory suggests that systematic bias in
attribution is less likely to occur in the fair value condition, and
likely explains the marginal signi?cance of the pooled results re-
ported in Table 2, Panel A.
H1b predicts that presenting fair values on the face of ?nan-
cial statements will mitigate the tendency to attribute exogenous
components of ?rm performance to managers and will improve
the accuracy of manager quality assessments. To test H1b, we es-
timate the regression used to test H1aand interact PosShock with
FV, where FV is equal to one (zero) if the measurement basis rec-
ognized on the face of the ?nancial statements is fair value (amor-
tized cost). If fair value presentation on the ?nancial statements
16
Because there is no normatively correct magnitude of participants’ quality rat-
ings, a ?nding that participant quality ratings are more positive (negative) for high
quality (low quality) managers when exogenous shocks are positive (negative) could
be consistent either with greater accuracy in assessing manager type, or with the
attribution bias we predict. However, in the case where there is a positive shock
and a low quality manager or a negative shock and a high quality manager (i.e.,
Cross = 1), participants’ manager quality ratings should not be associated with the
direction of the exogenous shock. As such, in our tests of H1a and H1b we focus
on instances where Cross = 1 and interpret an association between subject quality
rating and exogenous shock direction as a bias in attribution.
17
All reported p-values are one-tailed for directional predictions and two-tailed
otherwise.
attenuates attribution errors as predicted by H1b, we expect the
coe?cient on PosShock ? FV to be negative and signi?cant.
Panel B of Table 2 presents results that support H1b. As pre-
dicted, our results reveal that the coe?cient on PosShock ? FV
is negative (?26.18) and signi?cant (p = 0.001). Additionally, the
sum of the coe?cients for PosShock and PosShock ? FV are not
different from zero (untabulated, p = 0.12). Taken together, these
results suggest that recognition of fair value information miti-
gates systematic attribution bias and allows investors to disentan-
gle exogenous and endogenous factors when assessing stewardship
quality.
18
Test of H1b
The results of our ?rst set of hypotheses document the pro-
cessing differences investors use to reach stewardship assess-
ments when given ?nancial statements using differing measure-
ment bases. Speci?cally, we document the existence of an attribu-
tion bias in our setting and provide evidence that fair value recog-
nition ameliorates the bias. Our second set of hypotheses examines
whether the reduction of this bias through fair-value-based recog-
nition leads to more accurate manager quality assessments, and
whether the use of fair value (amortized cost) information in the
notes leads to more (less) accurate manager quality assessments.
For the remainder of our tests, we no longer restrict the dataset to
crossed ?rms, and each ?rm pair constitutes an observation in our
analyses.
18
As an alternate test of H1b, we estimate the regression restricting observations
to Cross = 0, and stack this regression with the tabulated regression in the same
variance–covariance matrix using seemingly unrelated regression techniques. This
speci?cation effectively tests a three-way interaction and allows us to empirically
test whether the difference in manager quality ratings between fair value and amor-
tized cost is driven by cases in which manager and shock types do not match. Unt-
abulated results support H1b (p < 0.01).
110 S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114
Table 2
Tests of H1a and H1b.
Predicted sign Coe?cient Standard error p-Value
Panel A: Regression examining effect of shock on manager quality assessments
MgrQuality = ?
0
+ ?
1
PosShock + b%
Constant 7.57
???
2.60 0.005
PosShock + 5.39
?
4.05 0.094
F-value 1.77
Number of observations 1424
Number of participants 73
Panel B: Effect of shock and measurement basis on manager quality assessments
MgrQuality = ?
0
+ ?
1
PosShock + ?
2
FV + ?
3
PosShock ? FV + b%
Constant ?2.28 3.18 0.476
PosShock 18.65
???
5.85 0.002
FV 19.51
???
4.69 0.001
PosShock ? FV ? ?26.18
???
7.54 0.001
F-value 6.55
Number of observations 1424
Number of participants 73
Notes: Panel A of this table presents the results of a regression model employed
to examine whether the direction of the shock affects participants’ ability to assess
manager quality after controlling for manager quality by restricting observations
to ?rms where manager quality and shock direction differ (i.e., Cross = 1). Panel B
presents the results of a regression model employed to examine whether the mea-
surement basis (either fair value or amortized cost) affects participants’ tendency to
rely on the shock to assess manager quality after restricting observations to ?rms
where manager and shock type differ. Each observation represents an individual
?rm manager. MgrQuality, FV, and PosShock are as de?ned in Table 1. p-Values are
estimated using Huber–White corrected standard errors clustered by participant.
???
Statistical signi?cance at the 1% level.
?
Statistical signi?cance at the 10% level.
To test H2a, we conduct a logistic regression with MgrCorrect
as the dependent variable and FV as the independent variable. The
higher quality manager in each ?rm pair is de?ned as the man-
ager who allocated a larger proportion of prior-period earnings to
the division with a higher net present value. In each ?rm pair, par-
ticipants rate both managers’ relative quality. Using participants’
ratings, our dependent variable MgrCorrect is equal to one (zero)
if a participant correctly (incorrectly) identi?es which ?rm had the
higher quality manager. We expect the coe?cient on FV to be pos-
itive and signi?cant.
Results are presented in Table 3, Panel A. Consistent with H2a,
the coe?cient on FV is positive (0.22) and marginally signi?cant (p
= 0.065).
19
This suggests that fair value recognition more transpar-
ently provides information that allows investors to correctly assess
manager quality relative to amortized cost recognition.
H2b (H2c) predicts that when participants assess stewardship
using ?nancial statements with recognized amortized cost (fair
value) information, the extent to which the participants utilize
supplemental fair value (amortized cost) footnote information will
improve (diminish) their ability to correctly identify higher qual-
ity managers. For H2b and H2c, we conduct a logistic regression
with MgrCorrect as the dependent variable and FV, NotesPercent,
and NotesPercent ? FV as independent variables. NotesPercent is the
percentage of total time that participants spend reviewing the al-
ternative cost basis notes to the ?nancial statements each round.
We expect the interaction coe?cient on NotesPercent ? FV to be
negative and signi?cant.
Table 3 presents the results of our test of H2b and H2c. Consis-
tent with our hypotheses, the coe?cient on NotesPercent ? FV is
negative (?6.80) and signi?cant (p < 0.001), suggesting that as the
19
When restricting our observations to ?rm pairs in which Cross = 1, the positive
coe?cient on FV is signi?cant at conventional levels (p = 0.03).
Fig. 4. Probability of investors identifying the higher quality manager. Note: Fig-
ure displays the probability that participants properly identi?ed the higher qual-
ity manager in our experiment given the percent of time they accessed the foot-
note information. Probabilities are tested for differences in Table 3, Panel B. FV (AC)
refers to conditions in which participants viewed fair value (amortized cost) ?nan-
cial statements with corresponding amortized cost (fair value) information in the
notes.
percentage of time viewing fair value (amortized cost) information
increases, stewardship assessment accuracy increases (decreases).
The signi?cantly positive coe?cient (1.11) on FV (p < 0.001) is
consistent with greater accuracy for participants in the fair value
condition after controlling for notes usage. Untabulated tests for
simple main effects also suggest that both fair value footnote use
and amortized cost footnote use contribute signi?cantly to the in-
teraction (p < 0.01).
To more completely test recognized fair values versus recog-
nized amortized cost with note disclosure of fair values, we com-
pare the probability that the participants identify the higher qual-
ity manager in (1) the case of fair value recognition with no
usage of amortized cost notes (i.e., NotesPercent = 0) versus (2)
the case of amortized cost recognition with the median amount
(10%) of fair value notes usage observed across participants in the
amortized-cost condition. Given these assumed levels of footnote
usage, we ?nd that participants in the fair value condition correctly
identi?ed the higher quality manager (68%) more often (untabu-
lated, p < 0.001) than participants in the amortized cost condition
(50%). These results provide additional support for the assertion
that amortized cost recognition with fair value disclosure is not a
substitute for fair value recognition. We also provide graphical ev-
idence of our interaction in Fig. 4 . As illustrated in Fig. 4, when
investors do not access the footnotes and just use the information
on the face of the ?nancial statements, investors make more ac-
curate stewardship decisions when fair value is recognized com-
pared to when amortized cost is recognized. However, the slope on
the regression line is negative for fair value and positive for amor-
tized cost. Consistent with expectations, the probability of identify-
ing the correct manager decreases as participants spend more time
in the footnotes that contain amortized cost information and in-
creases as participants spend more time in the footnotes that con-
tain fair value information.
Taken together, our tests of H2a–H2c suggest that the recogni-
tion of fair value information leads investors to make better stew-
ardship assessments, and that this result obtains because fair value
information is more transparently diagnostic than amortized cost
information. The actual use of fair value (amortized cost) informa-
tion, whether presented on the face of the ?nancial statements or
disclosed in the notes, leads to better (worse) stewardship assess-
ments. Moreover, disclosure of relevant fair value information is
not a substitute for recognition.
S.B. Anderson et al. / Accounting, Organizations and Society 46 (2015) 100–114 111
Table 3
Test of H2a, H2b, and H2c.
Predicted sign Relative risk coe?cient Standard error p-Value
Panel A: Effect of measurement basis recognition on participants’ stewardship decisions
P(MgrCorrect) = ?
0
+ ?
1
FV + b%
Constant 0.14
?
0.09 0.055
FV + 0.22
?
0.15 0.065
Wald Chi
2
2.28
Number of observations 1404
Number of participants 73
Panel B: Effect of measurement basis presentation and proportion of time spent accessing notes on participants’ stewardship decisions
P(MgrCorrect) = ?
0
+ ?
1
FV + ?
2
NotesPercent + ?
3
FV ? NotesPercent + b%
Constant ?0.37
???
0.15 0.011
FV + 1.11
???
0.22