Teaching

Accounting, Organizations and Society 27 (2002) 775–810
www.elsevier.com/locate/aos
Experimental research in financial accounting

Robert Libby *, Robert Bloomfield, Mark W. Nelson
Johnson Graduate School of Management, 383 Sage Hall,
Cornell University, Ithaca NY 14853 6201, USA
Abstract
This paper uses recent experimental studies of financial accounting to illustrate our view of how such experiments
can be conducted successfully. Rather than provide an exhaustive review of the literature, we focus on how particular
examples illustrate successful use of experiments to determine how, when and (ultimately) why important features of
financial accounting settings influence behavior. We first describe how changes in views of market efficiency, reliance
on the experimentalist’s comparative advantage, new theories, and a focus on key institutional features have allowed
researchers to overcome the criticisms of earlier financial accounting experiments. We then describe how specific
streams of experimental financial accounting research have addressed questions about financial communication
between managers, auditors, information intermediaries, and investors, and indicate how future research can extend
those streams. We focus particularly on (1) how managers and auditors report information; (2) how users of financial
information interpret those reports; (3) how individual decisions affect market behavior; and (4) how strategic inter-
actions between information reporters and users can affect market outcomes. Our examples include and integrate
experiments that fall into both the ‘‘behavioral’’ and ‘‘experimental economics’’ literatures in accounting. Finally, we
discuss how experiments can be designed to be both effective and efficient. # 2002 Elsevier Science Ltd. All rights
reserved.
1. Introduction number of experimental financial accounting

studies published in major accounting journals in
Financial accounting research is a broad field the 1960s and 1970s.
that examines financial communication between Serious criticisms of this early research (e.g.
managers, auditors, information intermediaries, Gonedes & Dopuch, 1974) turned experimentalists’
and investors, as well as the effects of regulatory focus away from financial accounting issues in the
regimes on that process. Much of this literature 1980s and early 1990s. As discussed by Maines
focuses on managers’ and auditors’ reporting deci- (1995) and Berg, Dickhaut, and McCabe (1995),
sions and their relationships to analysts’ forecasts major elements of these criticisms were: (1) the
and value estimates, investors’ trading decisions, irrelevance of individual behavior in market set-
and resulting market prices. This clear focus on tings, in which competitive forces will eliminate
judgment and decision making led to the large individual ‘‘errors’’; (2) poor matching of research
methods to research questions; (3) the lack of
* Corresponding author. Tel.: +1-607-255-3348; fax: +1- psychological or economic theory to predict effects
607-254-4590. and specify the mechanisms through which they
E-mail address: rl54@cornell.edu (R. Libby). occur; and (4) failure to capture relevant aspects
0361-3682/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved.
PII: S0361-3682(01)00011-3
776 R. Libby et al. / Accounting, Organizations and Society 27 (2002) 775–810
of the decisions of interest, in particular, decision economics’’ literatures in accounting.1 Although

maker attributes and institutional features. these literatures evolved from different traditions,
Beginning in the mid-1990s, there was a resur- we see them as essentially similar — both use
gence of experimental research addressing an even experiments to shed light on financial accounting
broader spectrum of financial accounting issues. issues, and therefore, both present similar oppor-
This paper presents our view of how this new lit- tunities and challenges to researchers. Naturally,
erature has addressed prior criticisms, and how it our review is also deeply affected by our own bia-
can continue to shed light on financial accounting ses and the financial accounting issues that we
questions. We argue that significant evidence of have been addressing in our own recent research.
capital market inefficiency has renewed interest in In Section 2, we describe in more detail how
how individuals make key accounting-related changes in views of market efficiency, reliance on
decisions and how these decisions affect market the experimentalist’s comparative advantage, new
prices. Recent studies take advantage of the theories, and a focus on key institutional features
experimentalist’s comparative advantage at disen- have allowed recent experiments in financial
tangling variables that are confounded in natural accounting to overcome the criticisms of the ear-
settings and measuring intervening processes to lier literature. In Section 3, we describe how spe-
draw strong causal inferences. Theories combining cific streams of experimental financial accounting
psychology and economics have allowed experi- research have addressed questions about financial
mentalists to specify more clearly the mechanisms communication between managers, auditors,
affecting individual and market behavior. Finally, information intermediaries, and investors, and
most of the new studies focus on issues of clear indicate how future research can extend those
relevance to financial accounting, particularly the streams. We focus particularly on (1) how man-
effects of decision-maker knowledge and motiva- agers and auditors report information; (2) how
tion, the complex information environment, reg- users of financial information interpret those
ulation, and strategic interaction. reports; (3) how individual decisions affect market
This paper is aimed primarily at those who plan behavior; and (4) how strategic interactions
to conduct financial accounting experiments, and between information reporters and users can affect
secondarily at other financial accountants who are market outcomes. While we address studies of
interested in what can be learned from experi- auditors in their financial reporting role, to limit
mental studies. Our primary goal is to use recent the scope of the review, we do not address issues
experimental studies of financial accounting to related to the demand for and conduct of auditing.
illustrate our view of how such experiments can be We also do not address studies of creditors’ deci-
conducted successfully. The core of our view is sions, which have received little attention in recent
that successful financial accounting experiments use financial accounting experiments.
the comparative advantages of the experimental In Section 4, we discuss how experiments can be
approach to determine how, when and (ultimately) designed to be both effective and efficient. We use
why important features of financial accounting set- the ‘‘predictive validity framework’’ (Libby, 1981;
tings influence behavior. By elaborating on this Runkel & McGrath, 1972) to structure our discus-
view, we hope to increase the impact of future sion of maximizing effectiveness through careful
experiments and help the new literature avoid the hypothesis development and research design. Our
mistakes and fate of the earlier literature. We do discussion of efficiency focuses on the consumption
not provide an exhaustive review of the literature, of scarce resources, such as subjects and compen-
nor do we provide detailed critiques of particular sation to those subjects. We conclude in Section 5
studies. Instead, we focus on how particular exam- with a brief summary of our main points.
ples illustrate successful use of experiments to
address important financial accounting issues. Our 1
See Haynes and Kachelmeier (1998) and Moser (1998) for
examples include and integrate experiments that recent discussions of the integration of the behavioral and eco-
fall into both the ‘‘behavioral’’ and ‘‘experimental nomic approaches to experimentation.
R. Libby et al. / Accounting, Organizations and Society 27 (2002) 775–810 777
2. Factors affecting the supply and demand for have no true economic effects (e.g. Andrade, 1999;
experimental financial accounting research Hand, 1990; Sloan, 1996; Vincent, 1997). Another
line of research indicates more generally that fun-
In this section, we examine four interdependent damental analysis of public financial statement
factors that have mitigated concerns raised about information can lead to higher stock returns (e.g.
the earlier experimental literature and promoted Frankel & Lee, 1998; Lee, Myers, & Swami-
recent progress in experimental financial accounting nathan, 1999; Ou & Penman, 1989). A third line of
research: changing views of market efficiency, research suggests that even sell-side analysts —
recognition of the strengths and weaknesses of generally recognized as among the most sophisti-
experimental methods in addressing financial cated users of financial statements — are pre-
accounting questions, the availability of new the- dictably biased (DeBondt & Thaler, 1990; Dechow
oretical bases for the research, and a more detailed & Sloan, 1997; La Porta, 1996).
view of the institutional features of financial The best-known lines of efficiency research focus
accounting settings. We discuss each of these factors on momentum in earnings and prices. A volumi-
in turn. nous literature on post-earnings-announcement
drift shows that markets underreact to large earn-
2.1. Changing views of market efficiency ings surprises (Ball & Bartov, 1996; Bernard &
Thomas, 1989, 1990; Bhushan, 1994; Brown & Han,
Much of the financial accounting research in the 2000; Foster, Olsen, & Shevlin, 1984). Another lit-
1960s implicitly assumed that some investors’ fail- erature, primarily published in finance journals,
ure to adjust fully for the effects of accounting shows that after adjusting for risk, stock returns
method choices would affect allocation of resour- are positively autocorrelated over periods of sev-
ces in the economy and disadvantage these less eral months (e.g. Chan, Jegadeesh, & Lakonishok,
sophisticated investors in their exchanges with 1996), but negatively autocorrelated over periods
more sophisticated investors (see Maines, 1995 for of several years (DeBondt & Thaler, 1985, 1987).
a review). A series of papers in finance (particularly The literature on market inefficiency is con-
Fama, 1970) persuaded many accounting research- troversial, and many of the papers alleging ineffi-
ers that if just a small fraction of investors are ciency have been criticized on methodological
sophisticated enough to respond appropriately to grounds (Ball, 1992; Fama, 1998; Kothari, 2000).
accounting information, they will compete among Nevertheless, many researchers now doubt whe-
themselves to set security prices equal to their ther markets satisfy the requirements of the semi-
expected values. As a result, the market becomes a strong form of the efficient markets hypothesis
‘‘fair game’’ in which even unsophisticated inves- (that markets respond efficiently to all publicly
tors are protected by the informational efficiency available information), or even the weak form
of prices.2 This research led Gonedes and Dopuch (that markets respond efficiently to information
(1974), among others, to argue that experimental contained in past market prices). Even some of the
research on individual behavior could have only most skeptical seem to be convinced that post-
limited importance for financial accounting. earnings-announcement drift is not simply an
In the late 1980s and 1990s, however, numerous artifact of research design (Ball, 1992). Recent
studies reported market inefficiencies.3 One line of research on efficiency has also led theorists to
research provides direct support for the assumptions examine how the assumptions underlying the effi-
underlying early financial accounting research: cient markets hypothesis might be relaxed to
accounting policies affect pricing, even when they account for archival results. (We discuss these
models more in Section 2.3). As a result, experi-
2
mental researchers can more easily argue that
Watts and Zimmerman (1986) also provided particularly
influential arguments.
individual behavior can be an important element
3
See Fama (1998), Kothari (2000), and Thaler (1999) for in determining market behavior, even in the pre-
more comprehensive reviews of this literature. sence of competitive forces.
2.2. The comparative advantage of financial experiments in financial accounting can rely on
accounting experiments well-developed psychological theories of judgment
and decision making4 that were in their infancy
Earlier financial accounting experiments typically when the studies reviewed by Gonedes and Dopuch
sought to determine whether specific accounting (1974) were conducted. Recent research can also
policy choices would affect investors’ decisions. rely on economic models that describe more care-
Answers to such research questions call for esti- fully when and how equilibrium outcomes arise.
mates of the magnitude of an effect (or error) by The major idea underlying much research on
representative actors in representative circum- judgment and decision making is that decision
stances, a task ill suited to experiments. Such a makers are boundedly rational (Simon, 1957).
task is more appropriate for archival-empirical Decision makers often have limited information
research, which examines large representative on which to base their judgments and decisions,
samples of naturally occurring phenomena. limited ability to retain and retrieve that informa-
More recent experimental research strives to use tion from memory, limited ability to process and
experimentalists’ comparative advantage to focus use that information, and limited insight into their
on disentangling the effects of variables that are own decision processes and future preferences.
confounded in natural settings and determining Studies over the last 25 years have focused on how
under what circumstances and through which various attributes of human cognition determine
processes specific phenomena arise. Experiments exactly what humans do well and what they do
are well suited to this task because they construct poorly. A number of their findings have influenced
their own research setting. In a constructed recent thinking in financial accounting and the
research setting, one can manipulate the indepen- study of financial markets.
dent variables, control for other potentially influ- Many decision-making studies emphasize the
ential variables by holding them constant or role of heuristics (Tversky & Kahneman, 1974).
through randomisation, and measure the inter- Heuristics are simplified decision rules developed
vening processes (such as information search or to deal with complex situations. These heuristics
the path players take to equilibrium outcomes in are efficient and often work well. But in some cir-
strategic settings) and mental states (such as cumstances they may lead to systematic biases
knowledge, beliefs, or confidence) that affect final such as over- and under-confidence in judgment
outcomes. This allows an experimentalist to dis- (Griffin & Tversky, 1992) and misperceptions of
entangle the effects of variables that are con- the covariation between signals and events (Lipe,
founded in the environment to draw strong causal 1991), which can systematically affect the manner
inferences, and to test the effects of conditions that in which individuals react to financial accounting
do not yet exist or do not exist in sufficient quan- information and the manner in which that infor-
tity in the natural environment (Libby & Luft, mation is impounded in prices. Learning to over-
1993). Experiments testing how and why (rather come biases is difficult because of the uncertainty
than whether or not) financial accounting phe- and poor feedback inherent in complex environ-
nomena occur can be based on theories of psy- ments. Often what we learn from experience is not
chological, economic or institutional processes. valid (Einhorn, 1980).
We discuss these theories next. The importance of (imperfect) storage and
retrieval of information from memory has also
2.3. Theoretical advances in psychology, finance, been recognized in recent financial accounting
and economics experiments. Some of these studies rely on models
Earlier experimental research was criticized for 4

Syntheses of the key constructs or ideas that drive psycho-
the lack of psychological or economic theory that logical theories of judgment and decision making have been
specified the mechanisms through which effects of provided by Carroll and Johnson (1990), Hogarth (1993),
accounting disclosures would occur. Recent Bazerman (1998), and others.
of memory organization (e.g. Smith & Medin, overconfidence into trading models. Other models
1981) that indicate how knowledgeable decision focus on forces that keep unbiased traders from
makers efficiently organize and retrieve data. exploiting price errors. For example, De Long,
Other studies recognize that memory for events is Shleifer, Summers, and Waldmann (1991) show
influenced by factors that are normatively rele- that traders who respond irrationally to irrelevant
vant, such as their frequency of occurrence, and information (‘‘sentiment’’) create enough noise in
factors that are normatively irrelevant, such as prices to keep rational traders from exploiting the
primacy, recency, and contrast effects (e.g. resulting price errors. Fischer and Verrecchia
Hogarth & Einhorn, 1992). Still others recognize (1999) and Kyle and Wang (1997) show that
that the limited capacity of working memory overconfidence, although irrational, can actually
affects our ability to consider multiple factors in give traders higher payoffs than their rational
making a judgment or choice. Consequently, even compatriots. These results make it difficult to
normatively relevant factors that decision makers argue that some form of natural selection will
are aware of often have limited influence on their eliminate irrational traders in dynamic equilibria,
judgments and decisions. and provide accounting researchers with specific
Recent research in accounting and finance also models of how and when individual biases might
relies on psychological models of risk (e.g. Kahne- influence market prices.
man & Tversky, 1979) and ambiguity (e.g. Einhorn Experiments focusing on game theoretic models
& Hogarth, 1986) that characterize individuals’ of financial accounting settings can now rely on
responses to risk and reward in ways that deviate new economic models that move beyond the tra-
from standard expected utility theory.5 This more ditional equilibrium view. Rather than simply
recent psychology literature provides greater abil- identifying an equilibrium and assuming that it
ity to predict under what circumstances behavior will occur, many economists have examined in
will be more or less likely to differ from the pre- detail what assumptions about rationality must be
dictions of standard economic theory (e.g. in satisfied for equilibria to have predictive power
earnings predictions versus trading behavior, in (Bernheim, 1984; Pearce, 1984; Tan & Werlang,
different information environments). A large lit- 1988). Other models have examined the process by
erature on social psychology could also be used to which equilibria are achieved, using either psy-
understand interaction between participants in chological theories based on behaviorism (Herrn-
financial accounting settings. For example, research stein & Vaughn, 1980) or evolutionary theories of
related to accountability (e.g. Tetlock, 1992), natural selection (Maynard Smith, 1982). In a simi-
motivated reasoning (e.g. Kunda, 1990) and group lar vein, Gode and Sunder (1993, 1997) used such
decision processes (e.g. Yetton & Bottger, 1982) ideas to show that ‘‘zero-intelligence’’ traders, who
has significantly influenced auditing studies. do nothing more than avoid obviously horrible
Other financial accounting studies use advances strategies, can achieve efficient security allocations
in financial economics to test the assertion that in some markets. By focusing on processes by
biased traders will be driven out of the market which equilibria are achieved, these studies provide
through systematic trading losses. Some of these indications of when equilibria will and will not
models focus on how biases might influence mar- predict behavior in financial accounting settings.
ket outcomes. For example, Barberis, Shleifer, and
Vishny (1998) use psychological models of how 2.4. Key institutional features of financial
people perceive random-walk sequences in a accounting settings
model with a representative investor. Daniel,
Hirshleifer, and Subrahmanyam (1998), Gervais Most early experimental studies in financial
and Odean (1997) and Odean (1998) incorporate accounting took relatively narrow views of finan-
cial accounting institutions. They typically focused
5
See Hodder, Koonce, and McAnally (2001) for further on the set of rules governing how accounting infor-
discussion of risk in financial accounting settings. mation could be reported in financial statements,
implicitly assuming that reporting choices (and Focusing explicitly on individual and environ-
interpretations of those choices) were made neu- mental characteristics allows experimental
trally, rather than being influenced by the incen- researchers to shed light on how and when
tives of a strategic manager or auditor. Early experimental results will generalize to target set-
studies also implicitly assumed that responses to tings, and also indicate how variations in these
financial accounting information would be inde- institutions will alter behavior. In this way, an
pendent of the expertise or incentives of the user, institutional focus helps researchers to exploit the
and that interactions among users and reporters comparative advantage of experimentation. In the
would not alter outcomes. next section, we describe how specific streams of
Consistent with the advice of Libby and Luft experimental financial accounting research have
(1993), recent experimental research in financial done so, and indicate how future research could
accounting has considered institutional features extend those streams.
more broadly, and has also focused on the inter-
action between individual and environmental
characteristics. Two key individual characteristics 3. Key financial accounting questions and
are the knowledge and motivation of information experimental evidence
reporters and users. These determine the parties’
goals, and how they use financial accounting to The goals of the literature that we review are
achieve those goals. Key environmental char- similar to those of the broader financial account-
acteristics include the complex regulations governing literature: to increase our understanding of the
ing reporting, the existence of financial markets, financial reporting process and its effects. While all
and the strategic interactions between reporters of the studies that we examine share the same
and users, as well as between different sets of general goal, they focus on different elements of
users. Regulations determine the set of choices the interactions of boundedly rational managers,
open to managers and auditors, and may also auditors, information intermediaries, and inves-
determine the results of those actions (e.g. lawsuit tors. These differences in emphasis led us to divide
outcomes). Financial markets affect how indivi- the studies into four related categories described
dual decisions result in aggregate market out- by the following questions.
comes, such as stock prices, liquidity and trading
volume, and may also determine wealth transfers
1. How do managers’ and auditors’ incentives
among different sets of investors. Strategic inter-
and financial accounting regulations deter-
actions capture the intertwining of the incentives
mine how they report events?
and actions of the many parties to financial
2. How do knowledge of accounting regula-
accounting decisions. Financial accounting set-
tions, managers’ incentives, and the infor-
tings include managers, auditors, investors and
mation content of accounting reports affect
information intermediaries (analysts and the
users’ (investors and information inter-
press) who may all interact strategically. Man-
mediaries) interpretations of accounting
agers and auditors negotiate to determine the
reports?
contents of the financial statement and audit
3. How do individual responses to information
report. Investors draw inferences about managers’
affect market-level phenomena?
and analysts’ information and incentives from
4. How do strategic interactions between
observing reports. Managers may choose reports
reporters and users of information affect
in an attempt to ‘‘fool’’ investors, but the investors
reporting and market outcomes?
may be able to anticipate these attempts.6
6
Financial accounting information is also used for con- We focus primarily on papers published since
tracting and stewardship purposes, but that has not been the the publication of Maines’s (1995) review of this
focus of significant experimental research. literature.
3.1. How do managers’ and auditors’ incentives and the circumstances under which they allow
and financial accounting regulations determine how managers to take aggressive accounting positions.
they report events? Consistent with the general auditing literature (e.g.
Kinney & Martin, 1994), results indicate that audi-
Reporting performance is fundamental to finan- tors reduce the aggressiveness of financial reports.
cial accounting. Discretion provided by financial For example, Hirst (1994) provides evidence that
accounting regulations, coupled with the inherent auditors consider management competence and
subjectivity of much accounting measurement, objectivity when evaluating management-provided
allows managers some flexibility to opportunisti- evidence. Phillips (1999) demonstrates that, after
cally report or manage earnings. Consequently, auditors receive evidence of aggressive reporting in
much archival and experimental research has high-risk accounts, they are more likely to attend
focused on this area. to it elsewhere, even in accounts they typically
Archival studies typically examine opportunistic consider to be of low risk. Kinney and Nelson
reporting by identifying whether earnings or (1996) demonstrate a circumstance in which
accruals differ from expectation in a manner favored auditors make audit-reporting judgments that
by managers’ incentives (see Healy & Wahlen, are as conservative as thought appropriate by
1999 for a review). While these studies have even those investors who are evaluating the
demonstrated numerous instances of apparent audit report in the presence of negative outcome
earnings management, their conclusions are some- information.
times criticized because of methodological diffi- However, other studies indicate that auditors
culties, including poor incentive proxies, misstated are more likely to allow their clients to take
discretionary accruals models, or potential omit- aggressive accounting positions when the relevant
ted variables such as operating choices that have evidence or precedents offer more room for inter-
non-earnings-management rationales but that pretation. For example, Nelson and Kinney (1997)
affect discretionary accruals (Bernard & Skinner, provide evidence that auditors are more (less)
1996; Dechow, Sloan, & Sweeney, 1995). Also, conservative than users required when the relevant
archival studies of earnings management focus on evidence was precise (ambiguous). Similarly, Salt-
post-audit financial statements that are a joint erio and Koonce (1997) provide evidence that
product of the negotiations between managers and auditors’ treatment of clients’ capitalization versus
auditors, which makes it difficult to distinguish the deferral decisions depends on whether the relevant
separate contributions of managers and auditors precedents unanimously favor one alternative.
to earnings management or to determine how When the precedents favor one alternative, audi-
managers’ and auditors’ separate incentives influ- tors follow the precedents, but when the pre-
ence their reporting and attesting behavior (Nel- cedents are mixed, auditors tend to follow their
son, Elliott, & Tarpley, 2000). client’s preference. Mayhew, Schatzberg, and Sev-
Experimental studies avoid these problems by cik (2000) provide consistent evidence in experi-
manipulating incentives and assessing treatment mental markets. When participants in the role of
effects rather than attempting to measure unex- auditor were sure of the appropriate disclosure,
pected accruals, and by holding constant task they made that disclosure, but as their uncertainty
characteristics that create potential omitted vari- about appropriate disclosure increased, they ten-
ables problems. Experiments can examine man- ded to misreport in favor of their client.
agers’ and auditors’ judgments separately, but can Other studies have focused on the role of spe-
also examine auditor–client interactions. These cific incentives in auditors’ reporting decisions.
characteristics of experimental work have led to a For example, Hackenbrack and Nelson (1996)
growing experimental literature that complements provide evidence that auditors are more likely to
the archival work in this area. allow their clients to take aggressive accounting
The largest group of experimental earnings- positions if the auditors’ litigation risk is reduced,
management studies focuses on auditors’ incentives and that auditors justify the aggressive position
with aggressive interpretations of the relevant managers of private firms, presumably because
financial accounting regulations. Hackenbrack managers of public firms face more disincentives
and Nelson hold constant the underlying audit for making income-decreasing financial accounting
evidence while varying auditors’ incentives and disclosures.
whether those incentives favored accrual or foot- The second approach is to elicit the joint pro-
note disclosure of a contingency, allowing them to duct of the manager–auditor negotiation indirectly
infer with high confidence that incentives were from auditors. Three different studies use different
driving the effects they observed. Using the same versions of this approach. Libby and Kinney
case materials, Kennedy, Kleinmuntz, and Peecher (2000) manipulate factors that affect managers’
(1997) provide evidence that, even when litigation incentives and ask auditors to determine how the
risk is relatively high, auditors may tend to take audited financial statements would appear. They
aggressive reporting positions when they can dif- provide evidence that correction of quantitatively
fuse personal responsibility by consulting other immaterial errors is much less likely if the correc-
experts within the firm. Wilks (2001) provides evi- tion would cause the firm to miss analysts’ EPS
dence that auditors’ interpretations of evidence forecasts (i.e. is qualitatively material), and that
and decisions are affected by the views of more the recently promulgated SAS 89 has little effect
senior auditors. Beeler and Hunton (2001) provide on this behavior. Gibbins, Salterio, and Webb
evidence that incentives from lowballing or man- (2000) develop a model of auditor–client negotia-
agement-advisory services affect audit partners’ tion and support their model by surveying audi-
going concern judgments. Bazerman, Morgan, tors concerning their experiences negotiating
and Loewenstein (1997) suggest that auditors contentious accounting issues with their clients.
cannot be independent because of the unconscious Nelson, Elliott, and Tarpley (2000) survey audi-
effect of such incentives, or even because of a sense tors concerning their experiences with clients’
of auditor–client affiliation that occurs through attempts to manage earnings, and provide evidence
multiple interactions. However, Dopuch and King concerning managers’ incentives for attempting
(1996) provide evidence that competitive pressures earnings management, the financial accounting
can reduce the effect of incentives like lowballing, areas in which managers attempt earnings man-
and King (2001) provides evidence that, holding agement, and the circumstances under which
constant economic incentives, professional–group auditors pass or thwart managers’ attempts.
affiliation can offset the influence of auditor–cli- Overall, these studies provide direct evidence
ent affiliation, demonstrating that offsetting that managers and auditors use the flexibility
affiliations can have offsetting effects on auditors’ inherent in accounting rules to make disclosures
independence. that are favored by their incentives. Holding con-
A smaller group of studies examines how man- stant amount of flexibility, changes in incentives
agers’ incentives affect the aggressiveness of their move disclosure in the direction favored by those
reporting decisions. These studies take two incentives. Holding incentives constant, increasing
approaches. One approach is to elicit managers’ flexibility increases the degree to which incentives
judgments directly. For example, Cloyd, Pratt, affect decisions.
and Stock (1996) gather data from corporate Certainly one direction for future research is to
financial executives at both public and private continue examining how managers’ and auditors’
manufacturing firms. They provide evidence incentives affect their decisions. In addition, the
that, when a manager has selected an aggressive literature could work more to identify the pro-
tax treatment, the manager tends to choose a cesses through which these effects occur. To what
financial accounting method that conforms to extent are these effects intentional and strategic
the tax choice in hopes of better defending the versus the unintended results of cognitive limita-
appropriateness of the tax choice if it is later tions? Wilks (2001) provides evidence that incen-
questioned by the IRS. Managers of public firms tives affect decisions more when the incentives are
were less likely to choose conformity than were made apparent to subjects prior to evaluating
evidence, suggesting that incentive effects influence 3.2. How do information users interpret reports,
the evaluation process as well as the decisions that given their knowledge of the regulations governing
result from that process. Beeler and Hunton those reports, and their knowledge of the reporters’
(2001) provide evidence that incentives affect both incentives?
the favorability and weighting of evidence, and
that auditors believe that incentives affect other Three streams of literature address distinct
auditors’ judgments, but not their own. A fruitful facets of this question:
direction for future research is to further under-
1. How do accounting methods and disclosure
stand how and when such incentive effects occur.
alternatives affect earnings predictions and
Another useful direction is to examine how
value estimates of investors and information
changes in regulations or other interventions
intermediaries?
might affect the aggressiveness of financial report-
2. How do investors and analysts use the time-
ing. For example, Libby and Kinney (2000), Hirst
series properties of earnings to predict future
and Hopkins (1998), and Maines and McDaniel
earnings?
(2000) provide evidence of recent regulatory
3. What determines analysts’ forecasting and
changes that do not appear to prevent managers
valuation performance?
from making aggressive reporting decisions. Cuc-
cia, Hackenbrack, and Nelson (1995) provide
evidence in a tax context that increasing the pre- We discuss each in turn.
cision of a standard does not prevent aggressive
reporting when the underlying evidence also pro- 3.2.1. How do accounting methods and disclosure
vides latitude for interpretation. When coupled alternatives affect earnings predictions and value
with evidence of the effect of incentives on report- estimates of investors and information intermediaries?
ing judgments, findings indicating the ineffective- The earliest experimental research in financial
ness of some regulatory interventions suggest that accounting tended to be motivated by the need for
regulators might reduce aggressiveness more evidence to address specific accounting policy
effectively by addressing incentives directly via debates. These studies focused on whether inves-
changes in penalties. Alternatively, other approa- tors and others adjusted appropriately for the
ches like improvements in audit-evidence sequen- effects of accounting methods and disclosure
cing (Phillips, 1999) or within-firm consultation alternatives (e.g. Dyckman, 1964; Jensen, 1966).
(Kennedy et al., 1998) might also affect the Looking back on the earlier literature, it is readily
aggressiveness of financial reports, by affecting the apparent that the answer to this question is
extent to which auditors discourage aggressive ‘‘sometimes.’’ Some participants in nearly every
reporting. study of this type demonstrate some degree of
Finally, future research could focus more on the functional fixation; they do not fully adjust for
interaction among participants in the financial differences in the effects of accounting alternatives
reporting process. Researchers are only beginning on the bottom line (Maines, 1995, p. 90, 91). As a
to consider the process by which auditors negoti- consequence, firms that are in identical economic
ate with their clients to produce the joint product circumstances except for their choice of accounting
that investors consume. Also, the increasing role alternatives are sometimes judged to be different.
of audit committees in this process remains lar- These specific policy-oriented studies did little to
gely uninvestigated. Addressing these issues via tell us how the extent of functional fixation will
experiments (e.g. Libby & Kinney, 2000), surveys vary across types of decision makers or economic
(e.g. Gibbins et al., 2000; Nelson, Elliot, & Tarp- circumstances, or what psychological processes
ley, 2000), and laboratory markets (e.g. Mayhew underlie insufficient adjustments to accounting
et al., 2000) appear to be useful directions for policies. Consistent with this concern, much
future research. These issues are discussed more in recent research has heeded the advice of Maines
Section 3.4. (1994) to focus on the dimensions of disclosure,
environmental factors, and processes that deter- Existing studies have examined three dimensions
mine the degree to which appropriate adjustments of classification. Hopkins (1996) examined the
are made. In response to a recent call for more effects of classification of items on the right side of
specific policy-oriented experiments (Beresford, the balance sheet as debt, equity, or mezzanine
1994), Maines (1994) noted that ‘‘Psychological and financing on judgments of the stock price effects of
sociological research may be most productively used new financing. He found that experienced buy-side
to guide behavioral accounting research on gen- analysts who had knowledge of the differential
eral issues that underlie many different accounting stock price effect of debt and equity issuances
standards, rather than focusing on issues relevant found in financial economics research responded
to only one standard.’’ Understanding the effects to the issuance of hybrid securities based on their
of these general factors will dramatically broaden categorization. When the securities were classified
the relevance of this research. as mezzanine, for which the analysts had no well-
Three groups of studies demonstrate progressive defined category, they responded based on the
refinement in the manner in which this research attributes of the individual security. Similarly,
question has been addressed. The first group Hopkins, Houston, and Peters (2000) examined
focuses on the mechanisms through which place- issues related to categorization of costs as operat-
ment and classification of accounting disclosures ing expenses, one-time charges, or note disclosure.
affect the use and interpretation of the disclosures. Experienced buy-side analysts treated the account-
The second group explicitly or implicitly recog- ing acquisition premium in a merger in part based
nizes that managers issuing accounting reports on its classification. One-time charges and note
have their own strategic interests and will report disclosures were treated as less relevant to stock
opportunistically, and examines how users valuation than operating expenses. Finally, Hirst
respond to voluntary disclosures by managers. and Hopkins (1998) and Maines and McDaniel
The third recognizes that analysts respond to their (2000) examined whether placement of elements of
own strategic interests and examines how users comprehensive income on the income statement
respond to potential relationship induced bias in versus the statement of stockholders’ equity affec-
analysts’ reports. We discuss each in turn. ted the ability to detect earnings management and
changes in earnings volatility. Information placed
3.2.2. General issues underlying functional fixation on the income statement (the primary perfor-
The development of category structures in mance statement) was much more likely to be
memory plays a major role in allowing expert treated as relevant to future performance esti-
decision makers to respond effectively and effi- mates by the experienced analysts in Hirst and
ciently in complex decision environments. In these Hopkins (1998) as well as by the evening MBA
structures, attributes are associated with cate- students in Maines and McDaniel (2000).
gories as opposed to individual instances of the Maines and McDaniel (2000) also present the
category. An individual instance or event is then beginnings of a theory of format effects. Their
interpreted based in part on its category member- theory lists five factors that affect the degree to
ship. This allows for efficient and often effective which investors will rely on a particular disclosure
processing of attributes of the environment, but in assessments of corporate performance: place-
sometimes produces errors when the particular ment, labeling as income, linkage (to net income),
instance does not match the typical category isolation, and degree of aggregation. Such a the-
attributes well. A number of recent papers have ory holds the promise of allowing predictions of
recognized that classification issues like the effects beyond the scope of individual studies, as
assignment of a financial disclosure to a particular Maines (1994) recommends. Future research can
financial statement, to a specific subsection within refine and test the model in other circumstances.
a statement, or to the notes, will affect decision Other studies identify the stage in the decision
makers’ categorization of that disclosure and process where any failure to adjust for accounting
interpretation of its relevance and meaning. or disclosure differences occurs. Following prior
credit analysis and auditing research (e.g. Abdel- the naive investor, as well as efficiency concerns
Khalik & El-Sheshi, 1980; Bonner, 1990), Lipe and Hand’s (1990) suggestion of investor sophisti-
(1998) uses a series of debriefing questions to cation effects as a partial explanation for market
separate the effects of measurement from weight- inefficiencies. This study, Maines, McDaniel, and
ing. She examines whether investors can accurately Harris’s (1997) study of segment standards, and a
assess the variance and covariance of returns in number of the above-mentioned are motivated in
making risk assessments and whether they use part by a particular policy issue of current interest.
those assessments in their investment decisions.7 Again, we believe that their impact is determined
Maines and McDaniel (2000) use a combination by their ability to relate the particular policy issue
of debriefing questions and regression analysis to of interest to more general phenomena that inform
determine whether differences in accessing the a wider array of policy questions.
information cues, interpreting or measuring the
cues, or weighting the cues caused their results. 3.2.3. Responses to voluntary disclosures
They suggest that participants in all disclosure The studies discussed above implicitly assume
conditions accessed and interpreted the cues in the that disclosures are generated by a neutral process.
same manner, but weighted them more heavily in However, managers issuing accounting reports
the income statement presentation condition. generally have their own strategic interests and
Another set of studies uses improved theories of will report opportunistically. A number of studies
functional fixation to define ‘‘superior’’ disclosure address how this strategic element affects users’
methods. Early studies only determined if different decisions.
judgments or decisions are made and ignore the The first two studies examine the effects of the
issue of determining the superior disclosure form of disclosures. Kennedy, Mitchell, and Sefcik
method. Many of the newer studies specify sub- (1998) examine how investors interpret the differ-
tasks necessary for successful final judgments or ent allowable forms of contingent environmental
decisions, such as detection of earnings manage- liability disclosure: minimum, best estimate, max-
ment (Hirst & Hopkins, 1998), assessment of imum, or range of the distribution. Experienced
variability in underlying ‘‘core’’ earnings (Maines financial executive, manager, banker, and MBA
& McDaniel, 2000), or covariance assessment (Lipe, student participants’ assessments of the distribu-
1998). Alternatively, Maines, Mautz, Wright, Gra- tion of possible losses implied by each disclosure
ham, Rosman, and Yardley (2000) approach the did not match the commonly accepted meaning of
question of assessing which disclosure method is the terms. For example, when the ‘‘best estimate’’
superior in a way similar to the training and deci- was disclosed by management, the participants
sion aids literature in auditing. They suggest that interpreted it as the minimum, and when a range
high quality reporting methods (1) allow novice was disclosed, the participants’ estimates of the
decision makers to perform like expert decision expected value were well above the midpoint of
makers and (2) allow the same decisions to be the range. The participants clearly believed that
made as completely disaggregated disclosures. managers bias their disclosures downward.8 It also
They apply their approach in a study of joint-ven- indicates that accounting information has different
ture financial reporting standards. The approach is effects on different judgments, in this case, man-
consistent with the SEC and FASB’s concern for agement credibility and firm value.
Hirst, Koonce, and Miller (1999) examine
7
She also examines how they react when market and investors’ interpretation of point versus range
accounting measures conflict. Her study is unique at this point forecasts and historic forecast accuracy on earnings
in jointly examining the role of accounting and non-accounting
8
information. It also suggests the possibility that the weight Participants also believed that managers that decided to
placed on normatively relevant information may change with disclose the minimum were the least credible, yet they valued
the inclusion of less-relevant information and presents a their firms the most highly. This suggests that the accounting
potential explanation for the lack of diversification of indivi- standard provides managers with a perverse incentive to pro-
dual portfolios. vide the least informative disclosure.
forecasts and confidence in forecasts (which they Similarly, the strength of the analysts’ arguments
relate to trading). If both of these forecast attri- had an effect only for negative recommendations.
butes indicate precision of the forecast, they both Ackert, Church, and Shehata (1997) extend this
should affect forecasts and confidence. However, study to a multiperiod setting where subjects have
only prior accuracy had an effect on earnings the option to acquire forecasts from analysts, and
forecasts, while both factors affected confidence also observe actual earnings. Individuals were
and trading. This again indicates that normatively much less willing to acquire analysts’ forecasts
relevant attributes of accounting information may that proved to be biased in the past, even when the
affect some judgments and decisions but not others. forecast information was useful. Both studies sug-
Libby and Tan (1999) and Tan, Libby, and gest the need to better understand the processes
Hunton (2000) investigate the effects of earnings that determine when reports from analysts and
warnings or preannouncements on sell-side ana- other information intermediaries will be purchased
lysts’ forecasts of future periods’ earnings. Libby and relied upon.
and Tan provide a demonstration of the process A general picture emerges from the above stud-
through which the same disclosure can have dif- ies. First, management’s often cited (Beresford,
ferential effects on different judgments and deci- 1994) preoccupation with the bottom line, and more
sions. They examine why analysts say in the press specifically with potential penalties for earnings
that they reward firms that warn, yet punish them volatility and effects of cosmetic differences,
in their forecasts. They demonstrate that this appears at least in part well founded. Second, we
inconsistency results from the simultaneous pro- have begun to understand that placement, categor-
cessing of the warning and earnings announce- ization, and labeling all play a role in the simplifi-
ment in answers to press questions versus the cations that even professional analysts apply when
sequential processing of the same signals in the evaluating accounting information. Future research
forecasting setting. Tan, Libby, and Hunton (2000) on the knowledge structures developed by experts
demonstrate that firms that low-ball preanno- for different types of companies and different types
uncements of both positive and negative earnings of financial judgments and decisions promises to
surprises will receive higher forecasts for future increase our understanding of these effects.
period’s earnings, even though the reporting man- It is also clear from the above results that the
agers themselves are judged as having lower information that decision makers rely upon in
integrity and competence. Also, analysts are aware their judgments is limited, and the information
of management’s tendency to low-ball the pre- emphasized clearly changes depending on the
announcements, but do not adjust their estimates financial judgment being made and other elements
of earnings of first time preannouncers in light of of the environment. In fact, awareness of cosmetic
this base rate knowledge. This again indicates that differences (and ability to ‘‘do the math’’) does not
known attributes of accounting information do ensure full consideration of their implications for
not affect all judgments in the same manner. valuation. The same is true of knowledge of man-
agement’s tendency to opportunistically employ
3.2.4. Responses to analyst’s forecasts vague reporting standards or analysts’ tendency to
Hirst, Koonce, and Simko (1995) and Ackert, bias their reports. There appear to be many cases
Church, and Shehata (1997) investigate the effects where the same normatively relevant factors are
of potential bias in analysts’ reports on investors’ ignored in one circumstance, but adequately
use of those reports. MBA student subjects in weighted in another by the same decision maker.
Hirst, Koonce, and Simko (1995) expected ana- The fact that results here tie closely to archival
lysts whose employers also provide investment data gathered in prior studies adds to the cred-
banking services to the company to be more ibility of the results. Future studies should focus
biased than those that do not. However, this per- on systematically determining the circumstances in
ceived bias only affected their reliance on the report which different classes of information receive first-
when the report gave a negative recommendation. order consideration.
Earlier research on the effect of task complexity replicate drift in the laboratory, using experi-
on the use of alternative decision rules in credit mental controls to rule out the possibility that
decisions (e.g. Biggs, Bedard, Gaber, & Linsmeier, prediction errors are driven by factors other than
1985; Paquette & Kida, 1988; see Payne, Bettman, judgment errors.9 Just as archival studies focus
& Johnson, 1992 for a review of psychological only on firms with extreme earnings surprises,
studies) will provide some guidance in this area. Calegari and Fargher use time series that exhibit
However, it appears that the determinants of which unusually large earnings changes in the most
information items receive first order consideration recent quarter. Their results are largely consistent
in particular judgment situations involves more with archival research — both individual traders
than task complexity. Findings of the importance and market prices underreact to earnings surprises.
of cue-response compatibility (Slovic & Lichten- Maines and Hand (1996) extend this finding in
stein, 1968) and other task determinants of cue two ways. First, they present MBA students with
usage in early judgment and decision making two different 40-quarter time-series. One series has
research (e.g. Einhorn & Hogarth, 1981; Slovic & strong autoregressive and moving-average com-
Lichtenstein, 1971) may provide useful directions ponents. Another is simply a seasonal random
for future research in this area. Furthermore, the walk with no such components. Subjects under-
interplay between these factors, investor sophistica- react to both elements when they are present, but
tion and effort, and various market attributes dis- also act as if the autoregressive element is present
cussed in Section 3.3 appear critical in determining when it is not. This suggests that drift may arise in
the importance of cosmetic disclosure differences. the target environment simply because it is too
difficult for investors to discern the autoregressive
3.2.5. How do investors and analysts use the time- and moving average terms. Drift may therefore be
series properties of earnings to predict future less severe for firms that adhere more closely to a
earnings? seasonal random walk. Second, Maines and Hand
Post-earnings-announcement drift has become a directly test Bernard’s (1993) hypothesis that
very active stream of archival research. Bernard investors anchor too strongly on earnings from the
and Thomas (1990) provide evidence that drift same quarter of the previous year, perhaps
arises because investors misperceive the time-series because it is stressed in the reporting format used
of earnings. Specifically, quarterly earnings follow a in the popular press. Maines and Hand test this
Brown–Rozeff model, which has two key elements. supposition by presenting a new set of subjects
One element is the autoregressive component — with a Brown–Rozeff time-series, and reporting
changes from one quarter of one year to the same earnings relative to earnings from four quarters
quarter of the next tend to be positively auto- ago. The results raise doubts about Bernard and
correlated. The other element is the ‘‘moving Thomas’s (1990) hypothesis, because these sub-
average’’ component — the differences between jects place even more weight on the autoregressive
actual and predicted earnings tend to be negatively component of the time series. These results suggest
correlated from one quarter to the same quarter of the need to test for alternative causes.
the next year. Research by Bernard and Thomas Bloomfield, Libby, and Nelson (2000a) argue
(1990) and Ball and Bartov (1996) indicate that that drift may arise because people naturally over-
investors underestimate both the autoregressive rely on unreliable information (Bloomfield, Libby,
and moving-average components of quarterly & Nelson, 2000b; Griffin & Tversky, 1992), and
earnings; results from Abarbanell and Bernard old earnings numbers tend to be unreliable pre-
(1992) indicate that analysts make a similar mistake. dictors of future earnings, once more current
Recent studies have used the advantages of the
9
For example, investors and analysts could appear to make
experimental approach to understand the psycho-
prediction errors in archival studies because they respond to
logical nature of investors’ and analysts’ time-ser- information other than earnings, because they have incentives
ies prediction errors. Calegari and Fargher (1997) for something other than prediction accuracy, or because they
provides a logical starting point — they attempt to are attempting to manage risk.
earnings are known. They test this hypothesis by 3.2.6. What personal and process attributes
manipulating information about old earnings per- determine analysts’ forecasting and valuation
formance, holding recent earnings performance performance?
constant. Student subjects rely much too heavily As Maines (1995) notes, a number of studies in
on old earnings numbers, and generate errors the 1970s and 1980s examined the manner in
consistent with post-earnings-announcement drift, which expert and novice analysts process
even when they are presented with a time series accounting information (e.g. Mear & Firth, 1987;
that is much simpler than that used in other Pankoff & Virgil, 1970; Slovic, Fleissner, & Bau-
experiments. This suggests that drift may not arise man, 1972; Wright, 1977). The studies assessed
merely because the time-series properties of earn- various characteristics of information search, cue
ings are so complex. weighting, judgment consistency and consensus,
Future research in time-series perceptions might and self-insight into information processing. A
follow several directions. One direction is to inte- number of the more recent studies in this group
grate the different research approaches described used detailed process tracing techniques in an
above. The realistic time-series used by Calegari attempt to tie individual or process attributes to
and Fargher (1997) and Maines and Hand (1996) judgment accuracy (e.g. Anderson, 1988; Biggs,
allow them to generalize their results readily to 1984; Bouwman, 1984). However, most studies
archival settings, but make it difficult for them to were only able to relate process attributes to
ascertain how aspects of the time-series data experience because of subject sample constraints
interact with psychological processes to cause or difficulty in measuring judgment performance.
prediction errors. The simpler time-series data These earlier experiments also did not focus on the
used in Bloomfield, Libby, and Nelson (2000a) effects of analysts’ incentives, which have received
poses precisely the opposite problem. Future a great deal of attention in recent archival studies.
research might attempt to work toward the middle Three recent studies have added substantially to
of these two approaches, either by using time- our understanding of the relationship of personal
series that are progressively simpler than in the and process variables to forecast accuracy as well
former studies, or progressively more complex as the impact of relationship incentives on bias in
than in the latter study. forecasts. Hunton and McEwen (1997) emphasize
Future research might also investigate the model both process measurement and disentangling
of Barberis, Shleifer, and Vishny (1998). That variables that are confounded in natural settings.
model assumes that earnings follow a random They address whether sell-side analysts’ search
walk, but that investors believe that earnings strategies and incentives (in the form of their rela-
switch between regimes of positive autocorrelation tionship to the company) affected the accuracy
and regimes of negative autocorrelation. This and bias of their earnings forecasts. Information
misperception results in both underreactions to search strategy was assessed with an eye move-
recent earnings changes and overreactions to long- ment measurement system that eliminates most
term trends. While such misperceptions are broadly concerns about the reactivity and validity of ver-
consistent with psychological findings indicating bal protocols. The authors measured the accuracy
representativeness and conservatism biases, no of the forecasts made in the experiment as well as
single study supports its assumptions, and their historical accuracy from company archives, which
predictions are not entirely consistent with archi- assures external validity. Analysts that followed a
val evidence (e.g. Lee and Swaminathan, 2000). more directed (as opposed to sequential) search
Finally, future studies might attempt to inte- strategies were more accurate both in the experi-
grate research on time-series predictions with mental task and in practice. The analysts in the
other research streams that consider earnings pre- underwriting condition gave higher (more biased)
diction more broadly. For example, how might forecasts than those in the following condition,
knowledge of earnings components (accruals, cash which were higher than those in the no relation-
flows) alter subjects’ time-series predictions? ship condition. Careful use of controls eliminates
concerns about omitted variables such as informa- would affect market-level prices in some straight-
tion availability, time on task, and some forms of forward manner (e.g. the price might be simply the
selection that could have explained similar findings average of all investors’ beliefs), and that some
in archival studies (see Kothari, 2000 for a review). investors would lose money to more sophisticated
Few studies have examined the knowledge and investors by trading unwisely at market prices.
abilities that lead to successful performance by Counter-arguments by proponents of the efficient
analysts. Ghosh and Whitecotton (1997) present markets hypothesis have led many experimental
evidence that two standard psychometric measures researchers to make these assumptions explicit and
of information processing ability (perceptual abil- subject them to testing. We divide this literature
ity and tolerance for ambiguity) were correlated into three lines: those that address differences
with forecast accuracy. But, as in Hunton and between individual and aggregate behavior, infor-
McEwen (1997), experience was unrelated to mation aggregation, and excess trading volume.
accuracy. However, Whitecotton (1996) reports
that experienced analysts outperformed MBA 3.3.1. Differences between individual and
students, who outperformed undergraduate stu- aggregate behavior
dents, though the experienced analysts were the A number of papers examine whether or not
most over optimistic. individual responses to information extend to the
Like similar work in auditing, these findings are market level. Two papers examine whether indivi-
potentially relevant to the selection and training of dual responses to risk extend to the market level.
analysts, as well as the interpretation of their Coller (1996) shows that both individual traders
forecasts and reports. Again, the fact that results and market prices respond to uncertainty in public
here tie closely to archival data, gathered either in disclosures in a manner roughly consistent with
the same study in the case of Hunton and McE- Bayesian rationality. Bloomfield and Wilks (2000)
wen’s (1997) accuracy measures, or in prior studies show that, consistent with theoretical and archival
in the case of their incentives findings, adds to the work on disclosure, more accurate disclosures
credibility of the results. Recent archival studies increase individual and market prices relative to
by Mikhail, Walther, and Willis (1997), Clement expected values, and also increase individual and
(1999), and Jacob, Lys, and Neale (1999) have market liquidity. A larger number of papers show
documented differences in the experiences of more that biases in individual decisions result in biased
and less accurate analysts that may indicate direc- market prices as well. For example, Calegari and
tions for future research. In the auditing literature, Fargher (1997) show that post-earnings-announce-
expertise studies have refined such findings in studies ment drift persists in a double auction market, and
that specify the knowledge necessary to complete Bloomfield, Libby, and Nelson (2000a) show that
various tasks, when it is acquired, and the over-reliance on previous years’ earnings persists
mechanisms through which knowledge content and in a clearinghouse market. Tuttle, Coller, and
structure affect performance. These studies can Burton (1997) show that recency effects extend to
provide guidance for future financial accounting the market level.
research in this area. Other recent work has begun Dietrich, Kachelmeier, Kleinmuntz, and Lins-
to look at how these individual responses affect meier (2000) conduct a study closely related to the
market-level performance and the characteristics functional fixation (e.g. Hopkins, 1996) and volun-
of markets that will affect information dissemina- tary disclosure (e.g. Kennedy et al., 1998) studies
tion. This research is discussed in the next section. discussed in Section 3.2.1. They demonstrate that
more explicit disclosure of accounting information
3.3. How do individual responses to information about oil-producing properties leads to more effi-
affect market-level phenomena? cient market prices even though the same infor-
mation can be inferred from the balance sheet and
Early experimental research in financial account- income statement. Different disclosure forms
ing implicitly assumed that individual behavior either mitigate or exacerbate biases in prices. The
authors test their process explanation by tying sively) than biased people, because biases are
individual participant’s behavior to prices to unconscious.
ensure that the market price results are the result Kachelmeier (1996a) uses an analysis of bids
of individual information processing biases. and asks to show the difficulty in determining
Other research investigates how competitive for- exactly how markets can debias prices. He induces
ces might allow less biased traders to have more a sunk-cost fallacy that significantly increases sell-
influence on price, and use that explanation to ers’ asking prices and buyers’ bidding prices.
guide examination of when this is more likely to However, these biases have no effect on transac-
occur. Of particular interest is the ‘‘smart-trader’’ tion prices, because the higher bids and asks cause
hypothesis, which states that traders who are less more trades to take place at the bids, which keeps
susceptible to the bias trade more actively than prices low.
other traders, driving prices to unbiased levels Other recent studies show that market structure
(Camerer, 1987, 1992). The intuition behind this can be important in determining when the smart-
hypothesis underlies the strong-form of the effi- trader hypothesis is likely to be supported.
cient markets hypothesis, which states that prices Ganguly, Kagel, and Moser (1994) present student
will fully reflect information even if it is held only subjects with a problem that leads to base-rate
by a small number of traders. neglect. They find that, because traders are not
Anderson and Sunder (1995) provide evidence allowed to sell shares they do not own (short-sell-
that the smart-trader hypothesis might be more ing is prohibited), market prices are set by the
predictive among professional traders than among traders with the highest valuation. As a result,
student traders. They compare the extent of base- market prices exhibit base-rate neglect most
rate neglect in markets involving student subjects strongly (weakly) when the biased prices are
with the bias in markets involving professional higher (lower) than the Bayesian expected values.
traders. They report that price biases in markets of Bloomfield and Wilks (2000) find strong indivi-
professional traders exhibit less base-rate neglect dual evidence of an ‘‘endowment’’ effect — incon-
over time, while price biases in markets of students sistent with Bayesian optimization, traders choose
do not. This is so even though the professional higher ask (selling) prices for riskier securities,
traders’ individual value estimates do not appear even as they simultaneously enter lower bid (buy-
to differ from the students’ estimates. This suggests ing) prices. However, higher risk does not cause
that the professional traders are able to trade in a the market ask price to rise. This form of irra-
way that reduces bias more (or increases it less). tionality at the individual level is eliminated at the
Bloomfield, Libby, and Nelson (1996) provide market level because the market ask is determined
evidence favoring the smart-trader hypothesis in a by the lowest individual ask. The market ask,
market in which security values are determined by therefore, reflects the selling price of the investor
the answer to general business knowledge ques- who succumbs least to the endowment effect. In
tions. Traders with more accurate answers do this way, the structure of the market combines
indeed trade more actively than other traders. with the nature of the bias to mitigate the bias at
When prices are influenced by trading volume, the market level.
prices become more accurate than the simple Future research could examine the foundations
average of all traders’ value estimates. (Prices are of the smart-trader hypothesis more directly. In
no more accurate than average estimates when particular, what factors might induce less-biased
they are not influenced by trading volume.) This traders to exploit biases, or keep them from doing
study might support the smart-trader hypothesis so? What factors might make more-biased traders
more strongly than the studies above because curtail their trading activity? How might changes
inaccurate traders are not biased, but merely in market structure, or the degree of market depth
uninformed. It is possible that uninformed peo- and liquidity, affect bias mitigation? (Archival
ple are more likely to know that their answers studies routinely show larger biases in less liquid
are inaccurate (and therefore trade less aggres- stocks.) Future research could also examine how
the nature of financial accounting information will The authors analyze the market’s reactions as the
affect the difference between individual and dividends are announced publicly one-by-one.
aggregate behavior. To the extent that informa- They find that the individual traders’ estimates of
tion induces biases, rather than degrees of value underreact slightly to the dividend announce-
informedness that differ across traders, prices ments, possibly because they erroneously believe
would seem more likely to represent an average of that random events tend to reverse over time (the
all traders’ beliefs. ‘‘gambler’s fallacy’’). More interesting is the fact
that market prices underreact substantially more
3.3.2. Information aggregation and underreaction than individual value estimates. The reason for
A different stream of research examines the this sluggishness in market prices is not clear, but
ability of financial markets to aggregate informa- the authors replicate it in both double-auctions
tion held by different traders. Like studies of the and call markets. Bloomfield, Libby, and Nelson
smart-trader hypothesis, aggregation studies are (2000b) also observe a similar effect in clearing-
motivated by the belief that traders who know a house markets. Bloomfield (1996a) shows that
security value does not reflect their own informa- markets react to a public signal when it is subject
tion will trade aggressively to exploit that fact, to manipulation by a self-interested seller, but not
thereby revealing their information to the market. when the signal is purely random. These results
Early studies on information aggregation raise the possibility that post-earnings-announce-
showed that markets do often aggregate informa- ment drift and underreactions to other informa-
tion. They do so most effectively when security tion (e.g. fundamental values, analysts’ estimates)
values are tied to states of nature in very simple may arise simply due to a generic underreaction of
ways (O’Brien & Srivastava, 1991; Plott & Sunder, market prices to information, rather than infor-
1988), and when experienced traders have com- mation-specific biases.
mon knowledge regarding the information envir- Several future directions for research in this area
onment (Forsythe & Lundholm, 1990). entail making endogenous the distribution of
More recent studies have examined how uncer- information among subjects. All of the aggrega-
tainty affects information aggregation. In a series tion studies described above manipulate informa-
of double-auction markets, Lundholm (1991) tion distribution by exogenously altering who is
manipulates the ‘‘aggregate uncertainty’’ that given information and who is not. Future studies
remains after combining investors’ information might relax this assumption by recognizing that
about security value. He finds that markets with collection of information is an intentional action
aggregate uncertainty aggregate information much that is driven in part by the perceived benefit of
less efficiently than those with aggregate certainty. becoming informed, as in Tucker (1997). Alter-
Imperfect aggregation can lead markets to under- natively, one might recognize that some informa-
react to information, because prices will be too tion may be effectively widely distributed because
high when the aggregate information indicates a it is more easily analyzed. For example, Sloan’s
very low value, and too low when the aggregate (1996) archival evidence that prices are too high
information indicates a very high value. Bloom- (low) when firms have high (low) accruals might
field (1996a, 1996b) shows a similar type of simply reflect an underreaction to financial state-
underreaction in a setting which allows aggregate ment information that is not widely known. This
certainty, but in which the information structure is result is consistent with Bloomfield and Libby’s
sufficiently complex that information aggregation (1996) finding that laboratory markets respond
is still very difficult. more strongly to information that is more widely
Other papers show that market prices can even available. However, a more direct test of this
underreact to public information that need not be hypothesis would be to give all traders the same
aggregated. Gillette, Stevens, Watts and Williams information (e.g. a complete financial statement),
(1999) construct a market in which security values and vary the ease with which the information can
are determined by a sequence of random dividends. be analyzed (as in Dietrich et al., 2000), as well as
the traders’ knowledge and training that would (although it has no apparent effect on price bia-
help with such analysis. ses). These results have regulatory implications:
More generally, researchers might start with the less sophisticated individual investors (who have
features we argue are essential for progress in less information than more sophisticated indivi-
functional fixation research — explicitly under- duals or institutional investors) can be protected
standing how people process and interpret the by regulations that emphasize the extent of their
information in financial statements, and then con- informational disadvantage.
sidering how differences in that processing might There appear to be a number of open questions
alter market behavior. related to trading volume. Archival papers have
examined volume in response to earnings
3.3.3. Trading volume announcements, or tie volume to pricing anoma-
A third line of research examines the determi- lies (Lee & Swaminathan, 2000; Swaminathan &
nants of trading volume in laboratory markets. Lee, 2000). These findings may be caused by fac-
Many of these studies are motivated by a general- tors indicated in economic models (e.g. Kim &
ization of the ‘‘no-trade’’ theorem (Milgrom & Verrecchia, 1994) or by psychological factors. The
Stokey, 1982), which shows that under fairly gen- literature on motivated reasoning seems particu-
eral conditions, information releases should not larly promising, because it examines how initial
induce any trading between traders. The intuition variations in beliefs and preferences can be mag-
is that if one trader expects to make money trad- nified by ambiguous public disclosures of infor-
ing at a given price, the trader on the other side of mation (Wilks, 2001).
the transaction must expect to lose money (since
trading is a zero-sum game). 3.4. How do strategic interactions between
Gillette et al. (1999) find routine violations of reporters and users of information affect reporting
the no-trade theorem: trading volume is generally and market outcomes?
quite high, and is even higher after very high or
low dividend announcements. These results are Game theory has been exceptionally useful in
consistent with archival evidence on trading modeling the strategic interactions between sellers
volume (e.g. Bamber, 1987; Bamber, Barron, & (who can make reports about their value) and
Stober, 1997), which have generated a number of buyers who rely on those reports in making their
theoretical models that generate trade through trading decisions. These models potentially have
complex interactions between public and private regulatory implications, because they show that
information (e.g. Kim & Verrecchia, 1994). How- seemingly reasonable regulations may be unneces-
ever, the simplicity of the market in Gillette et al. sary or unwise when one considers the joint
(1999) makes such explanations unlikely. response of buyers and sellers to the regulation.
Excess trading is a puzzle in Gillette et al. The models are very difficult to test with archival
(1999), but it has few welfare implications because methods, because their predictions are derived in
all traders are identical, and therefore wealth settings that are far simpler than natural markets.
transfers can be ignored (or are at best impossible However, a number of experimental researchers
to interpret). Bloomfield, Libby, and Nelson have chosen to examine behavior in settings that
(1999) examine excess trading that has very clear closely resemble those described in the models. In
welfare implications. They create markets in which this section, we briefly review some of these
less-informed traders hold a subset of the infor- experiments.
mation available to better-informed traders. Less- One line of research examines voluntary dis-
informed traders unwisely trade with — and lose closure models, in which sellers choose between
money to — the more-informed traders. However, honestly disclosing the exact value of the security
additional instructions that clarify to less- they are selling, and not disclosing anything at all.
informed investors the extent of their informa- Two papers by King and Wallin find strong sup-
tional disadvantages reduce these wealth transfers port for the qualitative predictions of the models
of Jung and Kwon (1988), and Wagenhofer ‘‘adaptive’’ strategies (doing more of strategies
(1990). King and Wallin (1991b) find that increas- that performed better in the past) lead to a given
ing the probability that the seller is informed leads equilibrium. For example, King and Wallin (1995)
sellers to disclose more often, and also leads buyers find little support for an ‘‘adaptively unstable’’
to draw more unfavorable inferences when they do equilibrium that is not the end result of adaptive
not observe disclosure (making disclosure a wise behavior. Other researchers focus more directly on
strategy for sellers). King and Wallin (1995) show the players’ thought processes. For example,
that disclosure is also limited by introducing a cost experiments by Bloomfield and Hales (2000)
to disclosing favorable information (a competitor examine how sellers’ abilities to form reputations
who will take advantage of favorable disclosures for honest reporting are influenced by buyers’ and
to enter the seller’s product market), because even sellers’ expectations of one another’s likely beha-
high-value firms might choose not to disclose. In vior and beliefs.
both cases, however, results deviate substantially Future research might also begin to integrate
from the point predictions of the models. disclosure research with the other literatures
Forsythe, Lundholm, and Reitz (1999) show described in this section. For example, Bloomfield
how disclosure regulations affect the welfare of (1996a) integrates the disclosure literature with the
buyers and sellers in a simple market with volun- information aggregation literature by showing that
tary disclosure. When sellers are not permitted to sellers are willing to pay a fee to inflate a public
disclose their information about value, many sur- signal, even though the information available to
plus-enhancing transactions do not occur, and the market as a whole is unchanged. They are will-
both buyers and sellers suffer. Allowing sellers to ing to do this because markets tend to react more
disclose any value (even a false one) increases strongly to information held by more investors.
market surplus, but these gains accrue almost Researchers might also integrate economics-
entirely to the sellers. Requiring sellers’ reports to based disclosure research with the psychology-based
include the true value shifts part of this surplus literature described in Section 3.1. That research
from the sellers to the buyers. focuses on how investors could use financial
King (1996) examines whether disclosure pat- reporting choices to draw inferences about man-
terns change when sellers have an opportunity to agers’ incentives and information, but ignores the
develop reputations. He permits sellers to report fact that managers should anticipate investors’
any value they wish, but imposes a cost on buyers reactions. On the other hand, the psychology-
when the seller’s report is inaccurate. This setting based research presents a more comprehensive
includes two equilibria. In an ‘‘inflation’’ equili- treatment of financial accounting institutions, by
brium, sellers always report the highest value, and allowing managers to choose how to classify and
buyers pay expected value net of the cost of inac- report accounting information. We believe it
curacy. In a ‘‘reputation’’ equilibrium, the seller would be worthwhile — though difficult — to
reports honestly, and the buyers believe the reports examine fully strategic interactions in more complex
until the seller reports dishonestly; at that point, accounting institutions. Researchers in financial
the players revert to the inflation equilibrium. accounting might also attempt to integrate game
King finds that an exogenous cost for inaccuracy theory and social psychology, as has been done
does permit reputation formation, but that the successfully in the auditing context by King (2001).
reputation equilibrium arises only in a few cases.
There are several natural directions for research
in strategic disclosure. There is certainly no short- 4. Effective and efficient research design:
age of new disclosure models to test. However, it is methodological considerations in experiments
probably more important for researchers to begin
to delve into how and why various equilibria do Section 3 presented a number of directions for
and do not have predictive power. Some research- future experiments. In this section, we discuss
ers have begun doing so by asking whether how these experiments can be designed to be
both efficient and effective. An experiment is effi- Fig. 1 illustrates the predictive validity model as
cient if it achieves a given level of effectiveness as it applies to Hypothesis H1b from Hunton and
economically as possible. An experiment is effec- McEwen (1997; hereafter, HM). As noted earlier,
tive if it provides evidence of sufficient internal based on prior theory and evidence HM hypothe-
validity that readers should believe the results of sized that sell-side analysts’ relationship-based
hypothesis tests, while being of sufficient external incentives would decrease their forecast accuracy.
validity that it bears on a significant part of the Analysts’ relationship-based incentives were oper-
financial accounting issue of interest.10 Both ationally defined as a three-level independent
internal and external validity are key to effective- variable: an ‘‘underwriting relationship’’ that has a
ness. An experiment that lacks internal validity direct impact on fees, a ‘‘following relationship’’
fails by providing a misleading indication of the that creates the need for future access to private
relation between the dependent and independent information, or ‘‘no future relationship.’’ HM
variable, while an experiment that lacks external expect analysts in the underwriting condition to
validity produces results that are (or at least provide the most optimistic forecasts, those who
should be) divorced from the motivation of the follow the firm to be next most optimistic, and
study. We do not provide an exhaustive treatment analysts who do not follow the firm to be the least
of research design (see Kinney, 1986; Runkel & optimistic. They operationally define optimism
McGrath, 1972; Trotman, 1996 for more compre- (the dependent variable) as the analysts’ forecast
hensive discussions). Rather, we focus on issues minus the actual earnings outcome. HM also con-
that we believe are particularly important or are trolled for a number of other potentially influential
often misunderstood. Section 4.1 addresses tech- variables including subject background, experi-
niques for maximizing effectiveness through care- ence, time on task, and information availability.
ful hypothesis development and research design. In Fig. 1, link 1 depicts the relationship in HM’s
Section 4.2 addresses when it is (and is not) possi- underlying theory. No theory can be tested
ble to improve efficiency by consuming fewer directly; rather, a theory is tested by assessing the
resources without sacrificing effectiveness. We relationship between the operational definitions of
address the number and type of subjects used in key concepts in the theory (i.e. by assessing link 4).
the experiment, the payment of monetary incen- For this test to be valid, the links between the
tives, the use of within-subject designs, and the concepts and the operational definitions (links 2
decision to use single-person tasks rather than and 3) must be valid, and other factors that might
interactive tasks (such as financial markets or affect the dependent variable (link 5) must be
strategic reporting settings). controlled or have no effect. A study’s internal and
external validity is determined by the validity of
4.1. Increasing experimental effectiveness these five links. We now discuss ways in which
researchers can strengthen each of these links.
We organize our discussion of experimental
effectiveness around the predictive validity model 4.1.1. Link 1: theory and hypotheses
(Libby, 1981; Runkel & McGrath, 1972). This The first determinant of experimental effective-
model provides a useful description of the ness is specification of a good research question. A
hypothesis testing process, and focuses our atten- good research question addresses the relation
tion on the key determinants of the internal and between two or more concepts, can be stated
external validity of a research design. clearly and unambiguously as a question, implies
the possibility of empirical testing, and is impor-
tant to the researcher and others (Kinney, 1986).
10
Internal validity is the degree to which you can be sure
Experimental tests of research questions must
that observed effects are the result of the independent variables.
External validity is the degree to which results can be general- rely on some theory depicting forces that influence
ized beyond the specific tasks, measurement methods, and par- behavior in the experimental setting. Theories may
ticipants employed in the study. range from highly specific numerical models (such
Fig. 1. Predictive validity framework.
as those derived from economics or artificial- empirically by extensions of the research that test
intelligence cognition models) to more general additional hypotheses concerning environmental
qualitative predictions based on prior evidence contingencies that define the limits of generality of
(such as systematic evidence that people use a the initial hypotheses (Trotman, 1996).
certain heuristic in a given setting). Regardless of For example, HM’s research question of ‘‘Do
its nature, the theory suggests the expected answer sell-side analysts’ relationships with the firms they
to the research question, and serves to guide the cover decrease their forecast accuracy?’’ relates an
many decisions and tradeoffs that must be made antecedent (analysts’ relationships) and consequ-
during the design and administration of an experience (forecast accuracy) that clearly maps into first
ment. Whereas archival researchers analyze data order concerns indicated by theory and prior evi-
from secondary sources,11 the experimental setting dence. If the experiment operationalizes those
is specifically designed to gather data relevant to concepts well and provides an internally valid test
the hypotheses. Consequently, all stages of the of their relation, it will provide insight into the
design of experiments are profoundly affected by real-world effect of analysts’ incentives on their
the need for a well-formulated research question judgments. Future research can then test the
and hypotheses. In this section, we emphasize four extent to which those insights can be generalized.
issues that are particularly important in develop- Second, experimental research questions in
ing good research questions and hypotheses in financial accounting should focus on how theories
experimental financial accounting research. drawn from fundamental disciplines (such as psy-
First, the hypotheses must have external valid- chology and economics) interact with details of
ity; that is, readers must believe that the theore- financial accounting institutions (as discussed in
tical concepts and the relationships between them Section 2.4). As Gibbins and Swieringa (1995)
capture important aspects of the target environ- suggest, accounting experiments should be ‘‘both
ment. Although people often speak of external theory driven and setting sensitive.’’
validity as an aspect of experimental stimuli, we Tying the accounting institution to theory from
consider it an element of theory as well. If the a fundamental discipline allows hypotheses to
theory and hypotheses are appropriately capturing have relevance beyond the very specific practice
relationships among elements of the target envir- context that motivated the experiment (as recom-
onment, an internally valid experiment will test mended by Maines, 1994). It also allows experi-
that theory in a manner that generalizes to the tar- menters to contribute to both financial accounting
get environment. External validity is established and the fundamental discipline. For example,
Nelson and Kinney (1997) apply Einhorn and
11
That is, the data is initially gathered for a different purpose. Hogarth’s (1986) ambiguity model to predict how
ambiguity affects financial statement auditors’ and Finally, experimental research questions should
users’ judgments of appropriate contingent-liabi- be based on a theory that describes causal rela-
lity disclosure. Their study shows how the differ- tionships between concepts. As discussed above,
ences between auditors’ and users’ incentives lead the key advantage of the experimental method lies
auditors to use the discretion provided by ambig- in its ability to disentangle factors that are con-
uous evidence to justify lower levels of disclosure founded in natural settings, and thus provide
than users desire. This result is of clear interest to indications of how and why phenomena arise. A
financial accounting researchers, and also con- causal theory also improves external validity,
tributes to psychologists’ understanding of how because causal forces are more likely to generalize
incentives interact with ambiguity.12 to different settings. This also leads to a preference
A more ambitious approach is to use funda- for research questions that focus on a directional
mental disciplines to develop and experimentally prediction of differences, as opposed to a single
test a general theory that is applied to the financial point prediction. As Trotman (1996) indicates,
accounting phenomenon of interest. For example, ‘‘the basis of any experimental design is that one
Maines and McDaniel (2000) identify various or more independent variables are manipulated
general dimensions of formats that signal infor- and the effect on the dependent variable(s) is
mation importance or that affect the cognitive cost observed.’’ Since experiments require abstraction
of processing information (see also Lipe, 1998). from the real world, any number of differences
They apply their theory when testing whether between the experimental and real-world environ-
information-disclosure format affects considera- ments could affect the particular levels of observed
tion of the volatility of unrealized gains and losses, measures. Consequently, evidence consistent with
but their theory is much broader than the parti- point predictions (e.g. ‘‘the market price will be
cular practice context that they examine. $5.00’’) and particular parameter estimates (e.g.
Third, researchers should frame their theories at ‘‘managers will weight current year’s earnings
the least specific level that can account for the data twice as heavily as prior year’s earnings’’) are
expected to arise from the experiment. Stating the unlikely to generalize to real-world environments.
theory with greater specificity will simply encou- Directional effects are more likely to generalize,
rage readers to argue that the results are driven by because differences between the experimental set-
a slightly different theory (such as a different the- ting and the target setting are more likely to alter
ory of categorization) that yields identical predic- the magnitude of an effect than its direction. A
tions in the experimental setting. Such debates are focus on directional effects also makes it much
rarely productive. If the distinction is likely to be easier to design an experiment that controls for
important in accounting settings, researchers inter- competing explanations. We discuss this latter
ested in accounting issues should consider what issue further in Section 4.1.3.
other experiments might illustrate this importance.
If the distinction is unlikely to have important 4.1.2. Links 2 and 3: operationalizing dependent
ramifications for accounting settings, experiments and independent variables
discriminating between such theories are more Link 2 relates the antecedent theoretical concept
appropriately seen as contributions to the funda- A to the independent variable(s) operationalized
mental disciplines from which the theory is drawn. in the experiment. Link 3 relates the consequential
concept B to the dependent variable operationalized
12
Of course, the theory should entail some element of doubt in the experiment. An internally valid test requires
before testing. Experiments applying psychology to accounting manipulation of each independent variable in a
settings can be uninteresting if readers are certain that the way that changes only one theoretical antecedent
results obtained in psychology will readily extend to accounting
at a time. At the same time, they must construct
even without seeing the experimental results. Experiments
applying economics to accounting settings can be uninteresting an operational dependent variable that measures
if they are little more than complex ways of showing that peo- the conceptual variable, and that variable alone.
ple prefer more money to less (Kachelmeier, 1996b). This section discusses three particularly difficult
issues in operationalizing variables: (1) choosing This discussion should not be construed as indi-
the appropriate realism of the stimuli presented to cating that all experiments testing theories drawn
participants; (2) choosing the appropriate levels of from psychology (economics) must have high (low)
independent variables; and (3) using measured stimulus realism. Experiments testing very general
independent variables. psychological theories (such as the relation between
short-term memory and optimism) could contribute
4.1.2.1. Realism of stimuli. A common challenge to financial accounting research with stimuli and
in operationalizing independent variables is decid- tasks that possess very low degrees of realism.
ing how realistic the stimuli should be. The Similarly, experiments testing the effects of super-
appropriate level of realism in the operationaliza- ior accounting knowledge on trading profits
tion of an independent variable is determined by would require high degrees of realism. It is the
the role of realism in the theory to be tested. goal of the experiment that determines whether
Experiments testing psychological theories typi- realism adds to or detracts from internal validity.
cally present participants with more realistic sti- Stimulus realism can also provide benefits
muli than experiments testing economic theory, beyond that required for an internally valid test of
because psychology-based experiments are typi- the underlying theory. First, realism can help
cally focused on how participants make decisions authors convey to readers the ways in which the
using cognitive processes and knowledge that results relate to prior research. For example,
developed in response to their real-world educa- Hopkins (1996) and Tan, Libby, and Hunton
tion, training, and experience. Without relatively (2000) are able to compare their pricing and earn-
realistic stimuli, participants may not rely on the ings-forecast difference results for some treatments
cognitive processes and knowledge of interest. For directly to prior archival studies, which increases
example, HM’s theory relates analysts’ knowledge confidence in the generality of the results of treat-
of their incentives to their earnings estimates. In ment combinations for which no (or insufficient)
order to test this theory, the experiment must archival data are available. Second, realism can
provide the participants with a sufficiently realistic help subjects understand the task they are being
stimulus to activate that knowledge. Similarly, asked to perform, thereby reducing noise in the
Hopkins (1996) tests the theory that classification data. This may be particularly important in eco-
of debt-equity hybrid securities alters analysts’ nomics-based experiments, which place high
inferences about firm value; this theory can be demands on participants’ attention.
tested only with relatively realistic stimuli and However, it is important not to exaggerate the
value-assessment tasks. benefits that stimulus realism provides when it is
Experiments testing economic theories typically not directly enhancing internal or external valid-
present participants with less rich information and ity. Such realism may not substantially increase
less realistic stimuli, because they focus on how par- external validity, which is determined mainly by
ticipants make decisions using economic informa- the theory itself and how effectively the theoretical
tion given particular preferences, constraints, and constructs have been operationalized. Similarly, it
incentives. The decision processes depicted in these is important not to exaggerate its costs. Experi-
theories are not hypothesized to depend on task mental economists often worry that realism may
realism, so these studies are less concerned with it. influence behavior in ways that lie outside their
For example, King and Wallin (1991a) test theories theories, and thus reduce internal validity
relating the probability that a seller knows the (Camerer, 1997; Smith, 1976), but as we will dis-
asset value to the sellers’ disclosure strategies and cuss in Section 4.1.3, these concerns typically can
buyers’ responses to those disclosures. That study be dealt with through good experimental design.
does not require realism or knowledge of parti-
cular real-world institutions, so it uses abstract 4.1.2.2. Choosing levels of independent variables.
stimuli and tasks to avoid introducing extraneous After choosing the nature of independent vari-
factors that might compromise internal validity. ables, the researchers must choose their levels. A
general goal is to choose levels that are different occurs prior to an earnings announcement, and
enough that the experiment has sufficient power to one in which the warning is followed by the nega-
yield strong effects, yet be within the relevant range. tive earnings announcement. A third condition
As indicated above, in some cases it is appro- cannot exist in practice: the warning and negative
priate to choose levels that depict real-world con- earnings announcement occur simultaneously.
ditions. For example, HM’s independent variable This ‘‘simultaneous warning’’ condition allows
consists of treatment levels that reflect what ana- them to separate the effect of the warning from the
lysts might experience in practice. Given that their sequential processing of two signals by creating
theory is testing the relation between those real- two comparisons (each treatment compared to the
world incentives and analysts’ behavior, this rea- simultaneous warning condition) that manipulate
listic depiction provides a strong test of the theory. only one antecedent. The other two settings
However, it is usually difficult to ensure a repre- enhance external validity by mapping naturally
sentative sample of independent variable values, into the institutional setting and archival findings
which limits the interpretability of levels of effects the authors seek to inform.
and parameter estimates in most experiments. Regardless of how one chooses the levels of the
Choosing realistic versions of naturally occurring independent variables, it is usually advisable to
phenomena can also make it difficult to manip- conduct manipulation checks. These are measures,
ulate only a single theoretical antecedent while often taken during debriefing, which seek to deter-
holding all others constant. This is particularly mine whether subjects noticed and interpreted cor-
true in studies of alternative accounting methods rectly the independent variable(s). Manipulation
or disclosures, where differences in method or dis- checks test link 2 of the predictive validity frame-
closure (the experimental treatments in these work. Manipulation checks are particularly useful
studies) can convey unintended information about when analyses reveal no significant treatment
the nature of the underlying transactions that effect, since one alternative explanation for the
affect the dependent variable but are not included lack of a significant effect is ineffective oper-
in the theory being tested. Experimental controls ationalization of the independent variable (a link 2
discussed under link 5 can be employed to reduce problem). However, it is critical that the manip-
this concern (e.g. Hopkins, 1996). ulation check tests recognition and comprehension
In other cases, it can be wise to create levels that of the independent variable, as opposed to serving
are unrealistically extreme. For example, For- as another test of the treatment effect. Otherwise,
sythe, Lundholm, and Reitz (1999) compare a the manipulation check is really just a second
regulatory regime that prohibits disclosure with measure of the dependent variable (testing link 4
one that allows any disclosure (even fraudulent rather than link 2).
statements). While these levels are unrealistic, they
allow a very powerful test of effects that would 4.1.2.3. Measured independent variables. Some
likely generalize to milder changes in disclosure independent variables in accounting experiments
regulations. are observed, rather than manipulated. Because
It can even be useful to specify at least one level subjects are not assigned randomly to measured
of the independent variables that cannot occur in treatment levels, measuring independent variables
practice, to enable a cleaner test of the underlying gives up some of the experimentalist’s comparative
theory. One example of this approach is provided advantage. Such studies are subject to the same
by Libby and Tan (1999). They seek to understand correlated-omitted variables problems that com-
how analysts can say they reward firms for issuing promise internal validity in archival research.
early warning of negative earnings surprises, while Therefore, it is typically preferable to manipulate
actually punishing them in their forecast revisions. important independent variables whenever possi-
Libby and Tan address this question by oper- ble, rather than measuring them.
ationalizing three ‘‘warning’’ conditions. Two However, there are at least four circumstances
conditions are realistic: one in which no warning where measuring independent variables is useful.
The first is that it is impossible or impractical to measure all of those forces simultaneously in one
manipulate an antecedent. For example, HM experiment. Thus, King measured some potential
hypothesize that analysts that are considered by intervening variables (he chose to examine how
their firms to be more accurate forecasters tend to sellers’ reporting accuracy affects buyers’ reliance
use a more directive, hypothesis-driven evidential on those reports), but not others.
search strategy. Because HM cannot randomly One way to avoid measured independent vari-
assign analysts to ‘‘high historical accuracy classi- ables is to construct separate experiments testing
fication’’ and ‘‘low historical accuracy classifica- the separate parts of the theory. Hopkins could
tion’’ treatments, it is possible that historic have tested the ‘‘a,b’’ and ‘‘b,c’’ links separately or
accuracy classification is correlated with some in sequence, reasoning that finding support for
other variable (such as age or intelligence) that both links suggests (but does not demonstrate) an
determines use of a directive search strategy. As a ‘‘a,c’’ link. However, he chose to provide a clean
consequence, HM include a number of control test of the ‘‘a,c’’ link by testing it directly, and
variables to test these alternative explanations for using subsequent measurement of ‘‘b’’ to provide
results, and are careful to discuss these results in comfort that subjects behaved as predicted. Simi-
terms of ‘‘associations’’ rather than ‘‘causes.’’ A larly, King could have separately tested buyers’
second reason to use measured independent vari- responses to seller decisions. However, we believe
ables is that the theory relating the antecedent to that both authors were justified in focusing their
the consequence involves mediating variables (a cleanest tests on the primary antecedent and con-
sequence of links through intervening variables). sequence concepts in their theory. A full under-
For example, Hopkins (1996) predicts that the bal- standing of the causal path may be somewhat
ance-sheet classification of manditorily redeemable encumbered by the problems associated with
preferred stock (concept a) affects analysts’ beliefs measured independent variables, but remaining
concerning the total amount of equity outstanding problems can be addressed in future research. For
(concept b), which in turn affects their stock price example, Bloomfield and Hales (2000) use a series
estimates (concept c). Because analysts’ beliefs of experiments to understand more of the linkages
about outstanding equity are actually a dependent in King’s study.
variable in a part of his theory (a affects b), Hop- Third, it is sometimes much less interesting to
kins cannot manipulate it directly. Those beliefs examine reactions to a manipulated variable than
become a measured independent variable when a naturally occurring one. For example, it would
testing the second part of the theory (b affects c). have been less interesting for Hopkins (1996) to
Similarly, almost every multi-person task involves test whether analysts who are told that there are
intervening variables, because the behavior of one more shares outstanding would place a lower value
person is determined by the (necessarily endogen- on a firm’s stock, all else held equal. It seems much
ous) behavior of another. For example, King (1996) more reasonable to ask whether the same analysts
tests whether imposing exogenous costs on buyers would use that belief to assess stock value when the
for inaccurate value estimates induces sellers to belief arises naturally. This type of concern is even
report values accurately. One simple breakdown more salient in tests of equilibrium models.
of this theory is that exogenous costs (concept a) Fourth, measured variables often provide the
reduce the prices buyers are willing to pay when keys to understanding underlying processes that
the seller has previously reported inaccurately produce the effects of interest. For example,
(concept b), which leads the seller to choose higher Maines and McDaniel (2000) make a contribution
reporting accuracy (concept c). Because equili- by demonstrating effects of format on judgments
brium models involve many forces acting simulta- of management effectiveness and stock risk (an
neously (e.g. the seller should anticipate the ‘‘a,b’’ link), even though their lack of significant
buyers’ response to his reports, and the buyers effects of format on valuation could be viewed as
should anticipate the seller’s response to their an insignificant ‘‘a,c’’ link. After all, each inter-
likely price-setting behavior), it is difficult to vening successive link adds noise and diminishes
the experimenter’s ability to detect an effect of ‘‘a’’ Very complex statistics are typically necessary in
on a later consequence (particularly when the later experiments only when they rely heavily on mea-
consequence is a very complicated judgment like sured independent variables, or when researchers
stock valuation). Only by eliciting intervening must try to boost power when subject resources
variables does a clear pattern of results emerge. are scarce. When those circumstances are not
Hirst, Koonce, and Miller (1999) demonstrate the apparent, complicated statistical tests may signal
importance of specifying the correct causal path. poor experimental design — the experimenter is
They show that the form of a forecast will affect trying to grapple after the fact with concerns that
trading decisions, not through estimates of future should have been headed off with good experi-
earnings, but through confidence in estimates. This mental design.
further highlights the need to elicit intervening This section describes some of the powerful
dependent variables that aid in interpreting results array of techniques experimenters can use to deal
with respect to tests of complex theories. We with extraneous variables. The most important
encourage researchers to measure potential inter- technique available to the experimentalist to con-
vening variables whenever possible, if only after trol for extraneous variables is to assign subjects
they measure their primary dependent variable.13 randomly to treatments. Random assignment,
combined with manipulation of independent vari-
4.1.3. Links 4 and 5: statistics and other ables, enables experimentalists to ensure that their
potentially influential variables results are not biased by factors of which they are
As noted earlier, internal validity refers to the aware, as well as factors of which they are not
degree to which variation in the dependent variable aware. For example, HM randomly assign ana-
can be attributed to variation in the independent lysts to incentive-treatment conditions. This
variable. Link 4 assesses the relations between the results in an unbiased distribution of industry
operational independent and dependent variables. familiarity, age, experience, prior accuracy, etc.
Link 5 captures ‘‘other potentially influential’’ or across the three levels of the incentive treatment.
‘‘extraneous’’ variables besides the independent Thus, HM can conclude, with a specified level of
variable that could affect the dependent variable. A statistical confidence, that these variables, and
key advantage of the experimental approach is that other unspecified variables such as motivation or
the effects of extraneous variables can be controlled breakfast size, did not account for the results. In
for primarily by holding them constant or through fact, had HM not chosen to measure analysts’
randomization. As a result, statistical analyses in experience and use it as a covariate to reduce var-
experiments are typically straightforward, often iance in their analysis, they could have ignored
consisting of simple t-tests, ANOVAs, or non- experience and expected that it would not affect
parametric equivalents. Extraneous variables can their mean results because of random assignment
also be measured as in archival studies, and used across treatment conditions.
to enhance the power of analyses by accounting More generally, random assignment to treat-
for variation in the dependent variable that is not ment conditions allows experimentalists to avoid
related to the theory being tested. Finally, extra- many of the omitted variable concerns that limit
neous variables can be manipulated to directly test causality inferences in archival studies. For exam-
their effect. Given the expense typically associated ple, Kothari (2000) notes that the direction of
with this approach, it should only be used when cause and effect between relationship and forecast
the experimenter believes the extraneous variables optimism documented in the archival literature is
cannot be dealt with another way. not clear. It could as easily result from managers’
selection of investment banks whose analysts pro-
13
Of course, the experimenter needs to worry about carry- vide a more optimistic forecast as from opportu-
over effects (i.e. earlier measurements affecting later behavior).
Sometimes the order in which successive dependent variables
nistic forecasting by analysts with relationships.
are elicited is manipulated between subjects to reduce this con- This selection alternative explanation is eliminated
cern. This is discussed further in Section 4.1.3. in HM by random assignment of analyst subjects.
As a result of random assignment, the expected tests whether subjects infer management signaling
value of analyst optimism prior to the treatment is or differential tax treatment from the balance
unbiased across the three incentive treatment sheet classification of the hybrid security. Either of
groups. these inferences could explain an effect of classifi-
Second, experimentalists can hold extraneous cation on forecast error, but neither is included in
variables constant at a particular level. For exam- Hopkins’ theory. Hopkins provides evidence
ple, HM hypothesize that analysts who will be against these explanations by eliciting in debriefing
underwriting securities exhibit different forecast subjects’ inferences about the underlying transac-
bias than analysts who do not, because they face tion and demonstrating a lack of significant dif-
different incentives. However, compared to non- ference in inference between treatment conditions.
underwriting analysts, underwriting analysts could Such measures operate much like a manipulation
also have larger amounts of information available check, but rather than providing evidence that the
about a firm, or spend different amounts of time independent variable operationalizes the ante-
forecasting earnings. HM deal with these potential cedent concept the experimenter intended, they
alternative explanations for changes in their provide evidence that the independent variable did
dependent variable (forecast accuracy) by holding not operationalize antecedent concepts other than
constant across treatments the amount of infor- those intended by the experimenter. The assurance
mation analysts have available and the amount of they provide is limited (in that they provide evi-
time analysts can spend on the experimental task. dence by finding an insignificant difference), but it
More generally, experimentalists typically hold is assurance nonetheless.
constant aspects of the institutional setting that A fourth way to deal with extraneous variables
they believe are potentially important but that are is to manipulate them and test their effects. For
not part of the portion of the research question example, Bloomfield, Libby, and Nelson (2000b)
examined in that particular study. present their subjects with a number of securities,
A third way to deal with extraneous variables is and vary between subjects the order in which
to measure them (typically during debriefing). securities are presented. They test for order effects
These measurements can be used as covariates or and find none, allowing them to discount order of
measured independent variables to account for presentation as a potential explanation for their
their effects. For example, HM identified prior results. Even if they did not test for such effects,
research that indicated that analysts’ forecast manipulating order in a balanced design would
accuracy changes as they become more experi- reduce the risk that results are specific to a parti-
enced. Since a general experience effect was not cular order. In general, manipulating factors
part of their hypotheses, but might affect their unrelated to the hypotheses can be useful, but
dependent variable, HM measured experience by expensive in terms of use of subjects.
eliciting years spent as a financial analyst and used Finally, experimentalists can deal with link-5
it as a covariate in their analysis. Years of experi- factors by ignoring them. By ‘‘ignore’’ we really
ence cannot have been influenced by HM’s treat- mean ‘‘abstract from,’’ because those factors will
ment effect, so they use it as a covariate to reduce not be included in the experimental environment.
noise in their analyses without fear that it is actu- Ignoring some extraneous variables is necessary
ally capturing some element of the effect of the because it is not practical to mimic all elements of
independent variable on the dependent variable reality in an experiment; some abstraction is
(link 4). Similarly, Hirst, Koonce, and Miller necessary for the experiment to be conducted in a
(1999) use a pretest measure of forecasted earnings timely manner. To the extent that subjects make
taken before the treatment was administered to assumptions about information that is not inclu-
reduce noise and increase power. ded in the experimental environment, those
Measurements of extraneous variables are also assumptions are randomly distributed across
useful for testing competing explanations for treatment conditions, and do not affect inter-
experimental results. For example, Hopkins (1996) pretation of results, as long as the treatments do
not differentially affect subjects’ assumptions those subjects with monetary incentives (which are
about extraneous variables. expensive); whether to use between-subjects designs
It is important to note that these methods of (which use more subjects than within-subjects
accounting for extraneous variables are effective designs); and whether to place subjects in a labora-
only when the experimental design manipulates the tory market (which requires more subjects than
variables of primary interest to test effects of would a study of individual judgments). Choosing
directional predictions. For example, Tuttle, Col- to consume more resources does not necessarily
ler, and Burton (1997) wish to examine how increase experimental effectiveness. Rather, it
security prices are influenced by the order in which increases effectiveness in some circumstances,
information is revealed to investors. They provide reduces it in others, and has a small enough effect
investors with rich firm-specific information about in others that it is not justified from a cost/benefit
market conditions and corporate events, rather perspective. We discuss each choice in turn.
than the abstract information used in many mar-
kets experiments. Because the authors cannot 4.2.1. Subject selection
know exactly what knowledge investors bring to When should experiments use professional sub-
bear in interpreting this rich information, it could jects? Our advice is to match subjects to the goals
have a number of unknown effects on stock price, of the experiment, but to avoid using more
and might lead prices on average to be higher and sophisticated subjects than is necessary to achieve
lower than they should be. However, rather than those goals.
comparing prices to a point prediction of true Experiments that examine the effects of some
value, they examine whether the order of infor- attribute subjects have developed before entering
mation release causes a difference in prices. This the experiment must use subjects who possess the
difference cannot be affected by extraneous vari- necessary attribute. Many studies use experiments
ables created by the rich information (although to ‘‘peer into the minds’’ of specific groups of
they surely exist), because the total information is experienced professionals to determine what they
held constant across the settings being compared. have learned about relevant concepts and events
As discussed by Bloomfield and Libby (1996), and how that learning affects decisions. Hopkins
this type of ‘‘paired securities’’ design can generally (1996) examines how knowledge of the differential
be used to eliminate concerns about unanticipated effects of debt and equity offerings determines how
effects of realism in experiments. Experiments that classification of debt-equity hybrids affects ana-
attempt to compare behavior to point predictions lysts’ judgments. Libby and Kinney (2000) seek to
sacrifice this powerful form of experimental con- explore how auditors’ beliefs about managers and
trol. Even apparently innocuous variables in an their own incentives determine the effect of old
experimental setting (such as the color of a com- and new regulations. In both of these cases, the
puter screen or the time of day at which data col- experimenter is interested in how subjects’ use of
lection occurs) could cause deviations of behavior some type of knowledge learned in the real world
from a point prediction, but are unlikely to cause causes treatment effects, so they must use subjects
those deviations to vary across levels of the with the requisite knowledge. Thus, these studies
manipulated independent variables. use professionals as subjects.
In some cases, the experimenter can train stu-
4.2. Increasing experimental efficiency (without dent subjects to possess an attribute (e.g. knowl-
compromising effectiveness) edge) that the experimenter is interested in
examining. This approach is cost-effective given
Experimenters make many choices that affect students’ greater availability than professional
the amount of resources consumed by their subjects, and is well suited for testing the effects of
experiments. This section discusses four such specific features of the learning environment and
choices: whether to use professional subjects elements of the resulting knowledge (cf. Bonner &
(which are difficult to obtain); whether to provide Walker, 1994). However, this must be done with
care since recently acquired knowledge is unlikely response to real world incentives to the experi-
to be of the same depth and breadth, or integrated ment. Such experiments attempt to examine how
as well with subjects’ pre-existing knowledge. professional practice has provided professionals
Student subjects are also entirely appropriate in with incentives that affect their behavior in parti-
studies that focus on general cognitive abilities, or cular ways. For example, HM studied the effect of
responses to economic institutions or financial- analysts’ incentives on their forecast accuracy,
market forces that are expected to be learned with those incentives determined by the analysts’
within the experimental setting. Maines and Hand perceptions and understanding of the relationship
(1996) provide an example of the former; they that the analyst has with the firm whose perfor-
examine the effects of general tendencies in the mance is being forecasted. Providing performance-
processing of time-series information on forecast- contingent incentives in this type of experiment
ing behavior. Any of the reporting studies by King would distort or interfere with the effects of the
and Wallin (1991a, 1991b, 1995) provide examples real world incentives, and is therefore inappropri-
of the latter; those studies examine how subjects ate. While the effects of professionals’ perceived
respond to the strategic forces in disclosure games. incentives might be diminished in the experimental
Other experiments focus on the judgments of setting, their direction should not be altered, so
general or novice investors, and so require subjects their directional effects should not be altered.
who possess only basic familiarity with accounting Experiments testing responses to economic the-
and investing. Student populations that have such ory (such as those described in Section 3.4) need to
basic familiarity are appropriate here as well. provide performance-contingent incentives in
MBA students and executive-program partici- order to induce subjects to possess the incentives
pants are particularly useful, as they often have assumed by the economic model (Smith, 1976).
some accounting knowledge and investing experi- Without such incentives, a fundamental causal
ence. Studies of this type employing student sub- element of the model may not be present, and
jects include Bloomfield, Libby, and Nelson (1999, there is no reason to expect theoretical predictions
2000a), Hirst, Koonce, and Miller (1999), Hirst, to hold. Performance-contingent incentives are
Koonce, and Simko (1995), Kennedy, Mitchell, almost always appropriate in laboratory market
and Sefcik (1998), Lipe (1998), Bloomfield and experiments that examine how individual biases
Libby (1996), Maines and McDaniel (2000), and can be mitigated by competitive forces. For
Nelson, Krische, and Bloomfield (2000). example, the ‘‘smart trader’’ hypothesis relies on
In general, experimenters should avoid using an assumption that more accurate traders trade
professional subjects unless it is necessary to more actively because they will earn money by
achieve their research goals. In addition to doing so.
increasing the experimenters’ own time and A researcher who has concluded that perfor-
expense, inappropriate use of professional subjects mance-contingent incentives are appropriate must
has negative externalities — they may make it then decide on how sensitive payments should be
more difficult for other experimenters to gain to variations in performance. Our casual observa-
access to this very valuable resource. tions suggest that most experimental tests of eco-
nomic theories pay subjects an average of $8 to
4.2.2. Monetary incentives $20 per hour, with payments ranging from $5/
When is it appropriate to provide explicit hour to $100/hour (or sometimes more). These
monetary incentives in financial accounting numbers reflect tradition and resource limitations
experiments? As in subject selection, the answer more than any reasoned theory. These incentives
should be driven primarily by the goals of the are obviously much less than most agents in
experiment. financial accounting target environments would
First, as noted above, experiments that focus on expect. However, we doubt behaviors would be
incentives rely on participating professionals to substantially different with larger incentives. Past
bring their knowledge of and behavior learned in experiments show little evidence that biases are
eliminated by incentive compensation, just as covariate in their analyses or by analyzing the

financial rewards have not allowed athletes to run change in forecasts caused by the treatment. Since
a 3-min mile. Limitations on abilities, rather than they want their subjects to attend carefully to the
a lack of reward, drive these results. More gen- information contained in their treatments, and
erally, larger monetary incentives might reduce the their analyses are based on comparisons between
size of biases, but are unlikely to alter their basic treatment conditions (which hold treatment sal-
nature and direction.14 Thus, larger incentives ience constant), they are not concerned about
would probably not change the inferences drawn drawing extra attention to the treatment.
from directional hypothesis tests. Within-subject treatments are particularly com-
mon in laboratory markets and games. For exam-
4.2.3. Within- vs. between-subjects designs ple, Bloomfield and Wilks (2000) create a setting
When should experiments use between-subjects in which each group of subjects participates in
designs, rather than within-subjects designs? eight different treatments (every cell of a 222
Within-subjects (or ‘‘repeated-measures’’ designs) design) over the course of two trading sessions.
where subjects provide more than one observa- Such repetition reduces noise in the data, which is
tion, generally enhance statistical power by allow- often high in early repetitions because the envir-
ing control of between-subjects differences (i.e. onment is so complex. Repetition also uses sub-
there is a ‘‘subject factor’’ in the analyses that jects’ time very efficiently, which reduces the
accounts for subject-specific noise). This approach already high cash cost of running such experi-
has the added advantage of using fewer subjects. ments. However, repetition also requires Bloom-
However, repeated measures designs can also field and Wilks to balance the orders of the
affect results by making treatment effects more treatments, to ensure that treatment effects are not
salient, which may signal to subjects that the confounded with order effects.
experimenter wants them to respond to the Tan, Libby, and Hunton (2000) also suggest the
manipulation (the familiar ‘‘demand effect’’ con- use of a combination of between- and within-sub-
cern). Also, repeated measures are vulnerable to jects designs as a method of partitioning the effects
carryover effects from the elicitation of one of unintentional biases from intentional judgment
measure to the next. Therefore, these designs are policies. Following Kahneman and Tversky (1996),
most effective when increased salience of manipu- they suggest that the between-subjects design pro-
lated variables is desirable from the standpoint of vides a clean test of the subject’s natural reasoning
the experiment’s goals and/or when any carryover process, while the within-subjects design draws
effect is desired or can be minimized via manip- attention to the independent variable of interest
ulation of the order in which measures occur. and thus gives the subject a chance to detect and
As noted earlier, Hirst, Koonce, and Miller correct errors and inconsistencies in their respon-
(1999) use one type of repeated measures design, ses. Comparison of results under the two approa-
the pretest–posttest design. Their subjects first ches highlights how subjects address any conflict
forecast earnings and assess confidence in that between what they do and what they know. Evi-
forecast, given only company background infor- dence of differences using between-subjects treat-
mation and the prior years’ financial data. The ments, but not using within-subjects treatments,
subjects were then provided with the experimental suggests that the between-subjects differences are
treatments (management forecast and information unintentional. On the other hand, evidence of dif-
about management forecast accuracy), and again ferences using within-subjects treatments, but not
forecasted earnings and assessed confidence. This using between-subjects treatments, suggests that
pretest–posttest design allows Hirst, Koonce, and subjects are aware of the implications of the differ-
Miller to increase power by using the pretest as a ences in the stimuli, but that, in their natural rea-
soning process, the stimuli were ignored or subjects’
14
See Kachelmeier and Shehata (1992) for a study on how related knowledge was not accessed and used. This
very large incentives influence responses to risk. method should be useful in other studies that
attempt to distinguish between the effects of judg- dramatically increase the cost of the experiment. A
ment heuristics versus knowledge. group of 50 subjects will yield 50 judgments that
The choice of between- versus within-subjects are statistically independent of one another.
designs affects analyses, since within-subjects mani- Forming those subjects into 10 separate five-trader
pulation (i.e. repeated measures) yields observa- markets yields only 10 judgments (market prices)
tions that are not independent. For example, that are statistically independent of one another.
Bloomfield and Wilks (2000) observe well over a As a result, the use of a market either reduces
thousand closing prices in their study. However, power or increases the costs of the study.
since there are only eight distinct groups of sub- Laboratory markets are most appropriate when
jects, their repeated-measures analyses effectively examining particular forces within the market that
compute the average treatment effect (a signed might affect bias mitigation (such as the smart-
difference) for each group, and then perform a t- trader hypothesis), or when examining dependent
test on the eight differences. This design is more variables that are simply undefined at the indivi-
powerful than it might seem, because each of the dual level (such as trading volume or market
eight numbers is the average of a large number of liquidity). Even in these cases, however, one can
observations, and therefore has very little noise. sometimes address experimental goals in indivi-
As discussed in Section 4.1.2, most laboratory dual decision-making tasks. For example, Nelson,
markets conduct supplementary analyses that Krische, and Bloomfield (2000) use an individual
break a theory into parts using measured inter- decision-making task to examine how confidence
vening variables. For example, Bloomfield and in one’s own ability to ‘‘pick winners,’’ relative to
Wilks (2000) examine how disclosure quality confidence in large-sample anomalies (such as
affects market price through its effects on market post-earnings-announcement drift) can affect tra-
liquidity, which is measured. It is more difficult to ders’ willingness to rely on a disciplined trading
apply pure repeated-measures statistical techni- strategy. They do not have traders transact with
ques to such analyses. However, experimenters each other, but rather examine the number of
should be aware that inappropriate statistical shares that each trader offers to transact. This
methods overstate sample size (and therefore approach allows researchers to examine the rela-
understate P-values), and should be interpreted tion between judgment and trading behavior, but
with caution. More importantly, researchers must does not allow researchers to capture strategic
make every attempt to use repeated-measures interactions between market participants.
analyses for their main hypothesis tests. Given that one chooses to conduct a financial
market, there are many decisions that can reduce
4.2.4. Using laboratory financial markets the cost of each observation. One method used
When is it necessary to place individuals in almost universally in laboratory financial markets
laboratory markets? Critics of individual decision- and laboratory games (as in Sections 3.3. and 3.4)
making experiments often suggest that biases and is to have each group provide many observations
suboptimal behavior would be driven away by (a repeated-measures design). As noted in Section
market forces. In our view, this criticism alone 4.2.3, repeated-measures designs offer many
rarely justifies the cost of a market experiment. As advantages, but affect the statistical analyses that
discussed in Section 3.3, few experiments have must be performed.
shown that market forces eliminate biases; even
when they mitigate a bias, they tend to affect its
magnitude, but not its sign (e.g. market prices are 5. Conclusions
still too high, but not by as much). Because only
directional effects are easily generalized from This paper discusses how recent experimental
experiments to target settings, using a financial research in financial accounting has responded to
market does not substantially alter an experiment’s past criticisms, discusses how the recent literature
effectiveness. On the other hand, the market does has developed and how it can be extended, and
provides our perspective on how future experiments overreaction/underreaction to earnings information as an

can be designed to maximize both effectiveness and explanation for anomalous stock price behavior. Journal of
efficiency. Our comments are driven by our belief Finance, 47(3), 1181–1208.
Abdel-Khalik, A. R., & El-Sheshi, K. (1980). Information
that experiments — whether based on psychologi- choice and cue utilization in an experiment on default pre-
cal or economic theory — must exploit the primary diction. Journal of Accounting Research, 18(2), 325–342.
advantages of the experimental method. Those Ackert, L. F., Church, B. K., & Shehata, M. (1997). An
advantages include the ability to construct an envir- experimental examination of the effects of forecast bias on
onment in which a causal theory of phenomena can individuals’ use of forecasted information. Journal of
Accounting Research, 35(1), 25–42.
be tested with a maximum of internal validity. Anderson, M. J. (1988). A comparative analysis of information
Experimental research is still only a small part and evaluation behavior of professional and non-profes-
of empirical financial accounting research. This sional financial analysts. Accounting, Organizations and
raises the question of how financial accounting Society, 13(5), 431–446.
Anderson, M. J., & Sunder, S. (1995). Professional traders as
experiments should relate to the more dominant
intuitive Bayesians. Organizational Behavior and Human
archival-empirical work. One of the most notable Decision Processes, 64(2), 185–202.
characteristics of the better studies that we have Andrade, G. (1999). Do appearances matter? The impact of EPS
reviewed is their close tie to formal or informal accretion and dilution on stock prices. Working Paper, Har-
empirical observation. These observations often vard Business School.
provide part of the motivation for the experi- Ball, R. (1992). The earnings–price anomaly. Journal of
Accounting and Economics, 15(2), 319–345.
mental studies, and are relied upon to demonstrate Ball, R., & Bartov, E. (1996). How naive is the stock market’s
the external validity of experimental results. use of earnings information? Journal of Accounting and Eco-
Future research can relate even more closely to nomics, 21(3), 319–337.
this literature by testing alternative potential Bamber, L. (1987). Unexpected earnings, firm size, and trading
explanations for archival findings when there are volume around quarterly earnings announcements. The
Accounting Review, 62(3), 510–532.
natural confounds, measurement problems, or Bamber, L., Barron, O., & Stober, T. (1997). Trading volume
where causality is unclear, by explaining contra- and different aspects of disagreement coincident with earn-
dictory findings, and by examining conditions ings announcements. The Accounting Review, 72(4), 575–597.
where large samples are unavailable. Experiments Barberis, N., Shleifer, A., & Vishny, R. (1998). A model of
can also point to directions for future archival- investor sentiment. Journal of Financial Economics, 49(3),
307–343.
empirical studies by specifying either the limits to Bazerman, M. H. (1998). Judgment in managerial decision
the generality of existing findings or other findings making. New York: John Wiley.
that should exist in further archival studies. Bazerman, M. H., Morgan, K. P., & Loewenstein, G. F. (1997).
The impossibility of auditor independence. Sloan Manage-
ment Review, Summer, 89–94.
Acknowledgements Beeler, J., & Hunton, J. E. (2001). Contingent economic rents:
insidious threats to auditor independence. Working Paper,
South Florida University.
Prepared for the Accounting, Organizations and Beresford, D. R. (1994). A request for more research to support
Society 25th Anniversary Conference, Oxford Uni- financial accounting standard setting AAA — accounting,
behavior and organization section. Behavioral Research in
versity, July 2000. We thank Ron King, Lisa
Accounting, 6(Supplement), 190–203.
Koonce, Laureen Maines, Greg Waymire, the par- Berg, J., Dickhaut, J., & McCabe, K. (1995). The individual
ticipants at the AOS 25th Anniversary Conference versus the aggregate. In R. H. Ashton, & A. H. Ashton
and the Emory Behavioral Financial Accounting (Eds.), Judgment and decision-making research in accounting
Research Conference for their comments and sug- and auditing (pp. 102–134). New York: Cambridge.
gestions, and Bernadine Low for her assistance. Bernard, V. L. (1993). Stock price reactions to earnings
announcements: a summary of recent anomalous evidence
and possible explanations. In R. Thaler (ed.), Advances in
behavioral finance (pp. 303–340).
References Bernard, V. L., & Skinner, D. J. (1996). What motivates man-
agers’ choice of discretionary accruals? Journal of Accounting
Abarbanell, J. S., & Bernard, V. L. (1992). Tests of analysts’ and Economics, 22(1–3), 313–325.
Bernard, V. L., & Thomas, J. (1989). Post-earnings announce- Calegari, M., & Fargher, N. L. (1997). Evidence that prices do
ment drift: delayed price response or risk premium? Journal not fully reflect the implications of current earnings for
of Accounting Research, 27(1), 1–48. future earnings: an experimental markets approach. Con-
Bernard, V. L., & Thomas, J. (1990). Evidence that stock prices temporary Accounting Research, 14(3), 397–433.
do not fully reflect the implications of current earnings for Camerer, C. (1987). Do biases in probability judgment matter
future earnings. Journal of Accounting and Economics, 13(4), in markets, experimental evidence. American Economic
305–340. Review, 77(5), 981–997.
Bernheim, D. (1984). Rationalizable strategic behavior. Econo- Camerer, C. (1992). The rationality of prices and volume in
metrica, 52(5), 1007–1028. experimental markets. Organizational Behavior and Human
Bhushan, R. (1994). An informational efficiency perspective on Decision Processes, 51(2), 237–272.
the post-earnings-announcement drift. Journal of Accounting Camerer, C. (1997). Rules for experimenting in psychology and
and Economics, 18(1), 45–65. economics, and why they differ. In Van Dam et al., Under-
Biggs, S. F. (1984). Financial analysts’ information search in standing strategic interaction: essays in honor of R Selten.
the assessment of corporate earning power. Accounting, Berlin, New York: Springer.
Organizations and Society, 9(3), 313–323. Carroll, J. S. & Johnson, E. (1990). Decision research: a field
Biggs, S. F., Bedard, J. C., Gaber, B. G., & Linsmeier, T. J. guide. Sage
(1985). The effects of task size and similarity on the decision Chan, L., Jegadeesh, K. C., & Lakonishok, J. (1996). Momen-
behavior of bank loan officers. Management Science, 31(8), tum strategies. Journal of Finance, 51(5), 1681–1713.
970–987. Clement, M. (1999). Analyst forecast accuracy: do ability,
Bloomfield, R. (1996a). The interdependence of reporting dis- resources, and portfolio complexity matter? Journal of
cretion and informational efficiency in laboratory markets. Accounting and Economics, 27(3), 285–303.
The Accounting Review, 71(4), 493–511. Cloyd, C. B., Pratt, J., & Stock, T. (1996). The use of financial
Bloomfield, R. (1996b). Quotes, prices and estimates of value in accounting choice to support aggressive tax positions: public
a laboratory market. Journal of Finance, 51(5), 1791–1808. and private firms. Journal of Accounting Research, 34(1), 23–43.
Bloomfield, R., & Hales, J. (2000). Developing reputations for Coller, M. (1996). Information, noise, and asset prices: an
reliable reporting: the role of expectations. Working Paper, experimental study. Review of Accounting Studies, 1, 35–50.
Cornell University. Cuccia, A. D., Hackenbrack, K., & Nelson, M. W. (1995). The
Bloomfield, R., & Libby, R. (1996). Market reactions to differ- ability of professional standards to mitigate aggressive
entially available information in the laboratory. Journal of reporting. The Accounting Review, 70(2), 227–248.
Accounting Research, 34(2), 183–207. Daniel, K., Hirshleifer, D., & Subrahmanyam, A. (1998).
Bloomfield, R., Libby, R., & Nelson, M. W. (1996). Commu- Investor psychology and security market under- and over-
nication of confidence as a determinant of group judgment reactions. Journal of Finance, 53(6), 1839–1885.
accuracy. Organizational Behavior and Human Decision Pro- DeBondt, W., & Thaler, R. (1985). Does the stock market
cesses, 68(3), 287–300. overreact. Journal of Finance, 40(3), 793–818.
Bloomfield, R., Libby, R., & Nelson, M. W. (1999). Confidence DeBondt, W., & Thaler, R. (1987). Further evidence of investor
and the welfare of less-informed investors. Accounting, overreaction and stock market seasonality. Journal of
Organizations and Society, 24(8), 623–647. Finance, 42(3), 557–581.
Bloomfield, R., Libby, R., & Nelson, M. W. (2000a). Over- DeBondt, W., & Thaler, R. (1990). Do security analysts over-
reliance on previous years’ earnings. Working Paper, Cornell react? American Economic Review, 80(2), 52–57.
University. Dechow, P., & Sloan, R. (1997). Returns to contrarian invest-
Bloomfield, R., Libby, R., & Nelson, M. W. (2000b). Under- ment strategies: tests of naive expectation hypotheses. Jour-
reactions, over-reactions, and moderated confidence. Journal nal of Financial Economics, 43(1), 3–27.
of Financial Markets, 3, 113–137. Dechow, P. M., Sloan, R. G., & Sweeney, A. P. (1995).
Bloomfield, R., & Wilks, T. J. (2000). Disclosure effects in the Detecting earnings management. The Accounting Review,
laboratory: liquidity, depth and the cost of capital. The 70(2), 193–225.
Accounting Review, 75(1), 13–42. De Long, J. B., Shleifer, A., Summers, L. H., & Waldmann,
Bonner, S. E. (1990). Experience effects in auditing: the role of R. J. (1991). The survival of noise traders in financial mar-
task-specific knowledge. The Accounting Review, 65(1), 72–92. kets. The Journal of Business, 64(1), 1–19.
Bonner, S. E., & Walker, P. L. (1994). The effects of instruction Dietrich, J. R., Kachelmeier, S. J., Kleinmuntz, D. N., & Lins-
and experience on the acquisition of auditing knowledge. The meier, T. J. (2000). Market efficiency, bounded rationality,
Accounting Review, 69(1), 157–178. and supplemental business reporting disclosures. Journal of
Bouwman, M. J. (1984). Expert vs. novice decision making in Accounting Research (in preparation).
accounting: a summary. Accounting, Organizations and Dopuch, N., & King, R. R. (1996). The effects of lowballing on
Society, 9(3), 325–327. audit quality: an experimental markets study. Journal of
Brown, L., & Han, J. (2000). Do stock prices reflect the impli- Accounting, Auditing and Finance, 11, 45–69.
cations of current earnings for future earnings for AR1 Dyckman, T. R. (1964). On the investment decision. The
firms? Journal of Accounting Research (in preparation). Accounting Review, 39(2), 285–295.
Einhorn, H. J. (1980). Learning From experience and sub- the determinants of confidence. Cognitive Psychology, 24(3),
optimal rules in decision making. In T. Wallsten (Ed.), Cog- 411–435.
nitive processes in choice and decision behavior (pp. 1–20). Hackenbrack, K., & Nelson, M. W. (1996). Auditors’ incen-
Hillsdale, NJ: Erlbaum. tives and their application of financial accounting standards.
Einhorn, H. J., & Hogarth, R. M. (1981). Behavioral decision The Accounting Review, 71(1), 43–59.
theory: processes of judgment and choice. Annual Review of Hand, J. (1990). A test of the extended functional fixation
Psychology, 32, 53–88. hypothesis. The Accounting Review, 65(4), 740–763.
Einhorn, H. J., & Hogarth, R. M. (1986). Decision making Haynes, C. M., & Kachelmeier, S. J. (1998). The effects of
under ambiguity. Journal of Business, 59(4), S225–S250. accounting contexts on accounting decisions: a synthesis of
Fama, E. F. (1970). Efficient capital markets: a review of theory cognitive and economic perspectives in accounting experi-
and empirical work. Journal of Finance, 25(2), 383–417. mentation. Journal of Accounting Literature, 17, 97–136.
Fama, E. F. (1998). Market efficiency, long-term returns, and beha- Healy, P. M., & Wahlen, J. M. (1999). A review of the earnings
vioral finance. Journal of Financial Economics, 49(3), 283–306. management literature and its implications for standard set-
Fischer, P., & Verrecchia, R. (1999). Public information and heur- ting. Accounting Horizons, 13(4), 365–383.
istic trade. Journal of Accounting and Economics, 27(1), 89–124. Herrnstein, R., & Vaughn, W. (1980). Melioration and behavioral
Forsythe, R., & Lundholm, R. (1990). Information aggregation allocation. In J. Staddon (Ed.), Limits to action: the allocation
in an experimental market. Econometrica, 58(2), 309–348. of individual behavior (pp. 143–176). New York, NY: Academic
Forsythe, R., Lundholm, R., & Reitz, T. (1999). Cheap talk, Press.
fraud and adverse selection in financial markets: some experi- Hirst, D. E. (1994). Auditor sensitivity to earnings manage-
mental evidence. Review of Financial Studies, 12, 518–581. ment. Contemporary Accounting Research, 11(1), 405–422.
Foster, G., Olsen, C., & Shevlin, T. (1984). Earnings releases, Hirst, D. E., & Hopkins, P. E. (1998). Comprehensive income
anomalies, and the behavior of security returns. The reporting and analysts’ valuation judgments. Journal of
Accounting Review, 59(4), 574–603. Accounting Research, 36(Supplement), 47–75.
Frankel, R., & Lee, C. (1998). Accounting valuation, market Hirst, D. E., Koonce, L., & Miller, J. (1999). The joint effect of
expectation, and cross-sectional stock returns. Journal of management’s prior forecast accuracy and the form of its
Accounting and Economics, 25(3), 283–319. financial forecasts on investor judgment. Journal of Account-
Ganguly, A. R., Kagel, J. H., & Moser, D. V. (1994). The ing Research, 37(Supplement), 101–124.
effects of biases in probability judgments on market prices. Hirst, D. E., Koonce, L., & Simko, P. J. (1995). Investor reac-
Accounting, Organizations and Society, 19(8), 675–700. tions to financial analysts’ research reports. Journal of
Gervais, S., & Odean, T. (1997). Learning to be overconfident. Accounting Research, 33(2), 335–351.
Unpublished Working Paper, University of Pennsylvania. Hodder, L., Koonce, L., & McAnally, M. L. (2001). SEC mar-
Ghosh, D., & Whitecotton, S. M. (1997). Some determinants of ket risk disclosures: implications for judgment and decision
analysts’ forecast accuracy. Behavioral Research in Account- making. Accounting Horizons (in preparation).
ing, 9(Supplement), 50–68. Hogarth, R. M. (1993). Accounting for decisions and decisions
Gibbins, M., Salterio, S., & Webb, A. (2000). Evidence about for accounting. Accounting, Organizations and Society, 18(5),
auditor–client management negotiation concerning client’s 407–424.
financial reporting. Journal of Accounting Research (in Hogarth, R. M., & Einhorn, H. J. (1992). Order affects in belief
preparation). updating: the belief-adjustment model. Cognitive Psychology,
Gibbins, M., & Swieringa, R. J. (1995). Twenty years of judgment 24(1), 1–55.
research in accounting and auditing. In R. H. Ashton, & A. H. Hopkins, P. E. (1996). The effect of financial statement classi-
Ashton (Eds.), Judgment and decision-making research in fication of hybrid financial instruments on financial analysts’
accounting and auditing (pp. 231–249). New York: Cambridge. stock price judgments. Journal of Accounting Research,
Gillette, A. B., Stevens, D. E., Watts, S. G., & Williams, A. W. 34(Supplement), 33–50.
(1999). Price and volume reactions to public information relea- Hopkins, P. E., Houston, R. W., & Peters, M. F. (2000). Pur-
ses: an experimental approach incorporating traders’ subjective chase, pooling, and equity analysts’ valuation judgments.
beliefs. Contemporary Accounting Research, 16(3), 437–479. The Accounting Review, 75(3), 257–281.
Gode, D., & Sunder, S. (1993). Allocative efficiency of markets Hunton, J. E., & McEwen, R. A. (1997). An assessment of the
with zero-intelligence traders: market as a partial substitute relation between analysts’ earnings forecast accuracy, moti-
for individual rationality. The Journal of Political Economy, vational incentives and cognitive information search strat-
101(1), 119–140(February). egy. The Accounting Review, 72(4), 497–515.
Gode, D., & Sunder, S. (1997). What makes markets allocationally Jacob, J., Lys, T., & Neale, M. (1999). Expertise in forecasting
efficient? The Quarterly Journal of Economics, 112(2), 603–630. performance of security analysts. Journal of Accounting and
Gonedes, N., & Dopuch, N. (1974). Capital market equilibrium, Economics, 28(1), 51–82.
information production, and selecting accounting techniques: Jensen, R. (1966). An experimental design for study of effects
theoretical framework and review of empirical work. Journal of accounting variations in decision making. Journal of
of Accounting Research, 12(Supplement), 48–129. Accounting Research, 4(2), 224–238.
Griffin, D., & Tversky, A. (1992). The weighing of evidence and Jung, W., & Kwon, Y. (1988). Disclosure when the market is
unsure of information endowment of managers. Journal of agreement to disagree. Can overconfidence survive the mar-
Accounting Research, 26(1), 146–153. ket test?. Journal of Finance, 52(5), 2073–2090.
Kachelmeier, S. (1996a). Do cosmetic reporting variations LaPorta, R. (1996). Expectations and the cross-section of stock
affect market behavior? A laboratory study of the accounting returns. Journal of Finance, 51(5), 1715–1742.
emphasis on unavoidable costs. Review of Accounting Stud- Lee, C., Myers, J., & Swaminathan, B. (1999). What is the intrin-
ies, 1, 115–140. sic value of the dow? Journal of Finance, 54(5), 1693–1741.
Kachelmeier, S. (1996b). Discussion of ‘‘tax advice and report- Lee, C., & Swaminathan, B. (2000). Price momentum and
ing under uncertainty: theory and experimental evidence.’’. trading volume. Journal of Finance (in preparation).
Contemporary Accounting Research, 13, 81–90. Libby, R. (1981). Accounting and human information processing:
Kachelmeier, S., & Shehata, M. (1992). Examining risk pre- theory and applications. Englewood Cliffs: Prentice-Hall.
ferences under high monetary incentives: experimental evi- Libby, R., & Kinney, W. R. (2000). Earnings management,
dence from the People’s Republic of China. The American audit differences, and analysts’ forecasts. The Accounting
Economic Review, 82, 1120–1141. Review (in preparation).
Kahneman, D., & Tversky, A. (1979). Prospect theory: an Libby, R., & Luft, J. (1993). Determinants of judgment per-
analysis of decision under risk. Econometrica, 47(2), 263–291. formance in accounting settings: ability, knowledge, motiva-
Kahneman, D., & Tversky, A. (1996). On the reality of cognition, and environment. Accounting, Organizations and
tive illusions. Psychological Review, 103(3), 582–588. Society, 18(5), 425–450.
Kennedy, J., Kleinmuntz, D. N., & Peecher, M. E. (1997). Libby, R., & Tan, H-T. (1999). Analysts’ reactions to warnings
Determinants of the justifiability of performance in ill-struc- of negative earnings surprises. Journal of Accounting
tured audit tasks. Journal of Accounting Research, 35(Sup- Research, 37(2), 415–436.
plement), 105–123. Lipe, M. G. (1991). Counterfactual reasoning as a framework for
Kennedy, J., Mitchell, T., & Sefcik, S. E. (1998). Disclosure of attribution theories. Psychological Bulletin, 109(3), 456–471.
contingent environmental liabilities: some unintended con- Lipe, M. G. (1998). Individual investors’ risk judgments and
sequences? Journal of Accounting Research, 36(Autumn), investment decisions: the impact of accounting and market
257–277. data. Accounting, Organizations and Society, 23(7), 625–640.
Kim, O., & Verrecchia, R. (1994). Market liquidity and volume Lundholm, R. J. (1991). What affects the efficiency of a mar-
around earnings announcements. Journal of Accounting and ket? Some answers from the laboratory. The Accounting
Economics, 17(1), 41–67. Review, 66(3), 486–515.
King, R. R. (1996). Reputation formation for reliable report- Maines, L. A. (1994). The role of behavioral accounting
ing: an experimental investigation. The Accounting Review, research in financial accounting standard setting. Behavioral
71(3), 375–396. Research in Accounting, 6(Supplement), 204–212.
King, R. R. (2001). An experimental investigation of self-serving Maines, L. A. (1995). Judgment and decision-making research
biases in an auditing trust game: the effect of group affiliation. in financial accounting: a review and analysis.
Working Paper, Washington University. In R. H. Ashton, & A.H Ashton (Eds.), Judgment and deci-
King, R. R., & Wallin, D. E. (1991a). Market-induced infor- sion-making research in accounting and auditing (pp. 76–101).
mation disclosures: an experimental markets investigation. New York: Cambridge.
Contemporary Accounting Research, 8(1), 170–197. Maines, L. A., & Hand, J. R. M. (1996). Individuals’ percep-
King, R. R., & Wallin, D. E. (1991b). Voluntary disclosures tions and misperceptions of time series properties of quar-
when seller’s level of information is unknown. Journal of terly earnings. The Accounting Review, 71(3), 317–336.
Accounting Research, 29(1), 96–108. Maines, L. A., Mautz, R. D., Wright, G. B., Graham, L. E., Ros-
King, R. R., & Wallin, D. E. (1995). Experimental tests of dis- man, A. J., & Yardley, J. A. (2000). Implications of international
closure with an opponent. Journal of Accounting and Eco- diversity in joint venture financial-reporting standards for finan-
nomics, 19(1), 139–168. cial analysts’ stock values. Working Paper, Indiana University.
Kinney, W. R. (1986). Empirical accounting research design for Maines, L. A., & McDaniel, L. S. (2000). Effects of compre-
PhD students. The Accounting Review, 61(2), 338–350. hensive income volatility on nonprofessional investors’ judg-
Kinney, W. R., & Martin, R. D. (1994). Does auditing reduce ments: the role of presentation format. The Accounting
bias in financial reporting? A review of audit-related adjust- Review (in preparation).
ment studies. Auditing: A Journal of Practice and Theory, Maines, L. A., McDaniel, L. S., & Harris, M. S. (1997). Impli-
13(1), 149–156. cations of proposed segment reporting standards for finan-
Kinney, W. R., & Nelson, M. W. (1996). Outcome information cial analysts’ investment decisions. Journal of Accounting
and the ‘expectations gap’: the case of loss contingencies. Research, 35(Supplement), 1–24.
Journal of Accounting Research, 34(2), 281–299. Mayhew, B. W., Schatzberg, J. W., & Sevcik, G. R. (2000). The
Kothari, S. P. (2000). Capital markets research in accounting. effect of accounting uncertainty and auditor reputation on
Journal of Accounting and Economics (in preparation). auditor independence. Working Paper, University of
Kunda, Z. (1990). The case for motivated reasoning. Psycholo- Wisconsin — Madison.
gical Bulletin, 108(3), 480–498. Maynard Smith, J. (1982). Evolution and the theory of games.
Kyle, A. S., & Wang, F. A. (1997). Speculation duopoly with Cambridge, UK: Cambridge University Press.
Mear, R., & Firth, M. (1987). Cue usage and self-insight of methodological perspective. Journal of Business, 45, 283–
financial analysts. The Accounting Review, 62(1), 176–182. 301.
Mikhail, M., Walther, B., & Willis, R. (1997). Do security Slovic, P., & Lichtenstein, S. C. (1968). The relative importance
analysts improve their performance with experience? Journal of probabilities and payoffs in risk taking. Journal of
of Accounting Research, 35(Supplement), 131–157. Experimental Psychology Monograph Supplement, 78.
Milgrom, P., & Stokey, N. (1982). Information, trade and com- Slovic, P., & Lichtenstein, S. C. (1971). Comparison of baye-
mon knowledge. Journal of Economic Theory, 26(1), 17–27. sian and regression approaches to the study of information
Moser, D. V. (1998). Using an experimental economics processing in judgment. Organizational Behavior and Human
approach in behavioral accounting research. Behavioral Performance, 6, 649–744.
Research in Accounting, 10(Supplement), 94–110. Smith, E. E., & Medin, D. L. (1981). Categories and concepts
Nelson, M. W., Elliott, J. A., & Tarpley, R. L. (2000). Where (pp. 1–17). Harvard: Cambridge.
do companies attempt earnings management, and when do Smith, V. (1976). Experimental economics: induced value the-
auditors prevent it? Working Paper, Cornell University. ory. American Economic Review, 66, 274–279.
Nelson, M. W., & Kinney, W. R. (1997). The effect of ambi- Swaminathan, B., & Lee, C. (2000). Do stock prices overreact to
guity on auditors’ loss contingency reporting judgments. The earnings news? Working Paper, Cornell University.
Accounting Review, 72(2), 257–274. Tan, H. T., Libby, R., & Hunton, J. (2000). Analysts’ reactions
Nelson, M. W., Krische, S. D., & Bloomfield, R. (2000). Sticking to earnings preannouncement strategies. Working Paper,
with the program: why investors don’t exploit anomalies shown Cornell University.
in large-sample studies. Working Paper, Cornell University. Tan, T. C., & Werlang, S. (1988). The bayesian foundations of
O’Brien, J., & Srivastava, S. (1991). Dynamic stock markets solution concepts of games. Journal of Economic Theory,
with multiple assets: an experimental analysis. Journal of 45(2), 379–391.
Finance, 46(5), 1811–1838. Tetlock, P. (1992). The impact of accountability on judgment and
Odean, T. (1998). Volume, volatility, price, and profit when all choice: toward a social contingency model. In L. Berkowitz (Ed.),
traders are above average. Journal of Finance, 53(6), 1887–1934. Advances in experimental social psychology 25 (pp. 331–376).
Ou, J., & Penman, S. (1989). Financial statement analysis and New York: Academic Press.
the prediction of stock returns. Journal of Accounting and Thaler, R. H. (1999). The end of behavioral finance. Financial
Economics, 11(4), 295–329. Analysts’ Journal, 55(November/December), 12–17.
Pankoff, L. D., & Virgil, R. L. (1970). Some preliminary find- Trotman, K. T. (1996). Research methods for judgment and
ings from a laboratory experiment on the usefulness of decision making studies in auditing. Melbourne, Australia:
financial accounting information to security analysts. Journal Coopers and Lybrand.
of Accounting Research, 8(Supplement), 1–48. Tucker, R. R. (1997). The relationship between public and pri-
Paquette, L., & Kida, T. (1988). The effect of decision strategy vate information: an experimental markets study. Behavioral
and task complexity on decision performance. Organizational Research in Accounting, 9, 219–249.
Behavior and Human Decision Processes, 41(1), 128–142. Tuttle, B., Coller, M., & Burton, F. G. (1997). An examination
Payne, J. W., Bettman, J. R., & Johnson, E. J. (1992). Beha- of market efficiency: information order effects in a laboratory
vioral decision research: a constructive processing perspec- market. Accounting, Organizations and Society, 22(1), 89–103.
tive. Annual Review of Psychology, 43, 87–131. Tversky, A., & Kahneman, D. (1974). Judgment under uncer-
Pearce, D. G. (1984). Rationalizable strategic behavior and the tainty: heuristics and biases. Science, 185, 1124–1131.
problem of perfection. Econometrica, 52(5), 1029–1050. Vincent, L. (1997). Equity valuation implications of purchase
Phillips, F. (1999). Auditor attention to and judgments of versus pooling accounting. The Journal of Financial State-
aggressive financial reporting. Journal of Accounting ment Analysis, 2(4), 5–19.
Research, 37(1), 167–189. Wagenhofer, A. (1990). Voluntary disclosure with a strategic
Plott, C., & Sunder, S. (1988). Rational expectations and the opponent. Journal of Accounting and Economics, 12(4), 341–363.
aggregation of diverse information in laboratory security Watts, R., & Zimmerman, J. (1986). Positive accounting
markets. Econometrica, 56(5), 1085–1118. research. Englewood Cliffs, NJ: Prentice Hall.
Runkel, P., & McGrath, J. (1972). Research on human behavior: Whitecotton, S. M. (1996). The effects of experience and con-
a systematic guide to method. New York: Holt, Rinehart and fidence on decision aid reliance: a causal model. Behavioral
Winston. Research in Accounting, 8, 194–216.
Salterio, S., & Koonce, L. (1997). The persuasiveness of audit Wilks, J. (2001). Predecisional distortion of evidence as a con-
evidence: the case of accounting policy decisions. Accounting, sequence of real-time audit review. Working Paper, Brigham
Organizations and Society, 22(6), 573–587. Young University.
Simon, H. A. (1957). Models of man. New York: Wiley. Wright, W. F. (1977). Financial information processing models:
Sloan, R. (1996). Do stock prices fully reflect information in an empirical study. The Accounting Review, 52(3), 676–689.
accruals and cash flows about future earnings. The Account- Yetton, P. W., & Bottger, P. C. (1982). Individual versus group
ing Review, 71(3), 289–315. problem solving: an empirical test of a best-member strategy.
Slovic, P., Fleissner, D., & Bauman, W. S. (1972). Analyzing Organizational Behavior and Human Decision Processes, 29,
the use of information in investment decision making: a 307–321.

Teaching

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Teaching

Încărcat de

Drepturi de autor:

Formate disponibile

Accounting, Organizations and Society 27 (2002) 775–810

Experimental research in ﬁnancial accounting

1. Introduction number of experimental ﬁnancial accounting

of the decisions of interest, in particular, decision economics’’ literatures in accounting.1 Although

Earlier experimental research was criticized for 4

Fig. 1. Predictive validity framework.

eliminated by incentive compensation, just as covariate in their analyses or by analyzing the

provides our perspective on how future experiments overreaction/underreaction to earnings information as an

S-ar putea să vă placă și