Description
Inthispaper,IprovideasummaryanddiscussionofChen,Rennekamp,andZhou(2015).Doingsoallowsmetheopportunitytodiscussthreeinterrelatedthemes–forecasting,motivatedreasoning,anddisag-gregation.Followingmysummary,Idiscussspecificopportunitiestoextendthislineofresearch.Ialsoofferexamplesofmoregeneralextensionsrelatedtothepsychologyandeconomicsofdisaggregation.
Accounting, Organizations and Society 46 (2015) 19–22
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
Discussion of c?The effects of forecast type and performance-based
incentives on the quality of management forecastsd?
R
Jeffrey Hales
Georgia Institute of Technology, United States
a r t i c l e i n f o
Article history:
Received 22 April 2015
Accepted 23 April 2015
Available online 5 May 2015
a b s t r a c t
In this paper, I provide a summary and discussion of Chen, Rennekamp, and Zhou (2015). Doing so allows
me the opportunity to discuss three interrelated themes – forecasting, motivated reasoning, and disag-
gregation. Following my summary, I discuss speci?c opportunities to extend this line of research. I also
offer examples of more general extensions related to the psychology and economics of disaggregation.
© 2015 Elsevier Ltd. All rights reserved.
1. Introduction
In discussing Chen, Rennekamp, and Zhou (2015, CRZ), I will
touch on three interrelated themes – forecasting, motivated rea-
soning, and disaggregation. To see how these themes relate, we
can start by viewing accounting through the lens of information
economics. In that light, accounting research can be characterized
as the study of how particular types of information (e.g., sales
forecasts, production estimates, external ?nancial reports, etc.) af-
fect the beliefs and actions of particular agents (e.g., managers, in-
vestors, creditors, auditors, etc.). I say affect, but the ?ow of causal-
ity is not unidirectional. As Christensen and Demski (2007) argue,
the information in accounting systems is endogenous because re-
quirements about what and how to measure can alter the supply
of transactions that will be observed. Thus, we cannot take infor-
mation as given, but rather must also understand how information
is created and utilized in equilibrium. Beliefs, therefore, matter and
motivated reasoning has much to say about how information will
be formulated and interpreted.
In a world with strict assumptions about individual rational-
ity and strong form market e?ciency, one could assume away
or marginalize many important accounting questions. However, as
these assumptions are relaxed, the issues of information produc-
tion and consumption become even more interesting as individual
sources of information can no longer be summarily replaced by a
single signal – market price – acting as a su?cient statistic for all
available information. Suddenly, ?nancial statements have meaning
and not just because they are used to calculate earnings. The com-
R
I thank Robert Bloom?eld, Shankar Venkataraman, and Donnie Young for help-
ful comments and Deloitte & Touche LLP for their generous funding of the 2014
Accounting, Organizations & Society Conference on Accounting Estimates.
E-mail address: [email protected]
ponents matter as well.
Within the ?rm, things are similarly complex and interesting.
Production managers make forecasts, superiors review budget re-
quests, and divisions work to meet performance targets. All of
these actions and decisions are based on data and convey infor-
mation. As noted by Butterworth (1972):
nderlying each standard input cost is an implied decision
about the kind and amount of resource to be used and the
nature of the market in which it is to be purchased. Behind
each input/output relationship is an implied decision regarding
the technical conditions under which the process should oper-
ate and the rate at which employees should work. In order to
obtain solutions to one set of problems – those for which the
output of a standard cost system will be considered useful – we
must assume the solution of many others. Often, our assump-
tions turn out to be wrong. The dilemma is universal and is
central to the problem of information system design.
p. 1–2
It is perhaps helpful to draw a distinction between information
and data. Data typically must be analyzed and aggregated in some
fashion to be used. In the words of Bloom?eld (2002), data must
be transformed into statistics to become informative. Nor is this
just a matter of practicality. While it might be infeasible today to
share every bit of business data captured by an enterprise resource
system with investors and creditors, that practical limitation will
someday be gone, if advancements in our ability to capture and
transfer data exceed advancements in our ability to generate data.
While advancements in the world of Big Data make it possible to
capture more and more bits of data, they do not eliminate the need
for aggregation and transformation.
1
1
While I’m referring speci?cally to aggregation and transformation here, I shouldhttp://dx.doi.org/10.1016/j.aos.2015.04.006
0361-3682/© 2015 Elsevier Ltd. All rights reserved.
20 J. Hales / Accounting, Organizations and Society 46 (2015) 19–22
Implicit in this drive to master Big Data is the idea that more
data is better, and often it is – especially when forming statistics.
But it need not always be the case that relying on ?ner parti-
tions of data will result in better statistics and/or improved deci-
sion making. Chen et al. (2015, CRZ) demonstrate one way in which
this could be true, but there are likely others, as I will discuss later.
The format of my discussion is as follows. In the next section,
I provide a summary of the ?ndings in CRZ. Following the sum-
mary, I discuss speci?c opportunities to extend CRZ before closing
with a discussion of more general extensions related to the idea of
disaggregation.
2. Summary of CRZ
CRZ conduct an experiment demonstrating the tradeoffs that
can arise when agents forecast components of a summary mea-
sure of their performance, rather than just directly forecasting the
summary measure itself. As the results of CRZ indicate, forecasting
components results in an improved forecast of the summary mea-
sure (i.e., it is more accurate and no more biased), but only when
agents do not have a monetary incentive linked to the performance
measure. When participants do have a monetary incentive linked
to the performance measure, disaggregated forecasting harms fore-
cast quality, resulting in forecasts that are no longer more accurate
and that are now biased.
Because of the context of their experiment, CRZ ?ts solidly into
the managerial accounting literature on budgeting and forecasting.
Research on management control systems has typically focused on
a single input/output for the control system. CRZ add additional
depth to this literature by noting that, in naturally occurring set-
tings within organizations, essentially every budget request, pro-
duction forecast, or sales estimate will be supported – explicitly
or implicitly – by underlying components of data, reminiscent of
Butterworth (1972). Management control systems could, therefore,
be designed to encourage – or discourage – explicit assessment of
these components.
If all agents were perfectly rational, this particular system de-
sign choice would have no effect on the outcome. But, intuitively,
it is easy to imagine that a control system could act as a decision
aid by encouraging agents to explicitly consider more components
than they might otherwise naturally be inclined to do. In other
words, introducing more decision/assessment points into the de-
cision making process, could combat an important potential short-
coming to optimal decision making – miserly cognition. But, in do-
ing so, it can also open the door to another potential problem –
increased bias in processing.
This is where, from a theoretical perspective, CRZ make another
contribution by extending prior work on motivated reasoning. Prior
accounting research has examined how motivated reasoning alters
investors’ processing of information (Hales, 2007) and subsequent
search (Thayer, 2011). More recently, researchers have also demon-
strated that motivated reasoning in?uences how investors inter-
pret certain types of quantitative management guidance (Han &
Tan, 2010) and certain types of non-quantitative language (Hales,
Kuang, & Venkataraman, 2011). CRZ extend this line of research by
adding what I view as an interesting innovation – examining the
multiple behavioral effects that can arise when a high-level judg-
ment is broken down into components.
In the experiment, CRZ manipulate the incentive schemes their
participants face, which they use to alter the strength of motivated
reasoning. This manipulation has theoretical value, by helping CRZ
also note that the linguistic ?eld of pragmatics has much to say on a related topic –
namely, conveying information through elevation and implicature, as discussed by
Bloom?eld (2012).
to isolate their constructs of interest, and also practical value be-
cause compensation schemes can be altered in practice. Thus, if
?rms need to elicit forecasts from individuals with strong direc-
tional incentives, they may want to encourage relatively high-level
assessments. In contrast, if the forecaster is expected to be rela-
tively objective (i.e., directional incentives tied to the forecast are
relatively neutral), encouraging the forecaster to explicitly assess
lower-level components might be more valuable.
3. Opportunities for extending CRZ
3.1. Understanding the process
CRZ posit that having additional assessments could hurt a bi-
ased forecaster by limiting the bene?t of random error cancelation.
In other words, while disaggregation could help an objective fore-
caster be more accurate (and no more/less biased), having a biased
forecaster explicitly assess components will cause the component
error terms to become correlated. As CRZ argue, the cost of cor-
related error terms could offset the bene?t of getting potentially
smaller error terms, resulting in an aggregate forecast that is more
biased and no more accurate than if the biased forecaster were
to have simply made a high-level forecast. CRZ provide some ev-
idence on this process in their supplemental analysis section, but
a future study could be designed to more carefully look at the un-
derlying process and to examine alternative forecasting techniques
that might result in low error and bias.
For example, instead of forecasting either a high-level fore-
cast or a set of component forecasts, individuals could be encour-
aged/required to do both. In that context, would order matter or
would a c?top-downd? approach and a c?bottom-upd? approach
end up at essentially the same spot? If it is the latter, would that
end result simply be in the middle of the two single step forecast-
ing conditions or would the second stages (having to forecast an
aggregate from components or components from a high-level fore-
cast) be differentially in?uential on the ?nal outcome?
3.2. Task settings
One conference participant raised the topic of multi-product di-
visions. In those settings, bottom-up forecasting might have a dif-
ferent impact as information makes its way up the chain. In par-
ticular, one difference is that, as pieces of information move up the
chain, the strength of the directional preference associated with
that information could weaken because the individuals using the
information as an input are less directly involved in the production
task being forecasted. In other words, we might not see the same
increase in optimism when one person is forecasting the perfor-
mance of other individuals compared to when he or she is break-
ing their own performance up into component pieces. In addition,
as forecasts are moving up the chain, there is always the possibil-
ity that the superior will anticipate some degree of bias and error.
Since bias is directional, it might be easier for a superior to undo
bias in a subordinate’s forecast than to ?lter out unsigned errors.
However, even unraveling anticipated bias can be a tricky prospect
(see, e.g., Cain, Loewenstein, & Moore, 2005).
During the conference, another participant noted that optimistic
bias about future performance is viewed as negative in CRZ, which
seems appropriate given that the task is to convey information.
But one could imagine that optimism might not be all bad. In
fact, research in psychology has documented a number of bene-
?ts associated with dispositional optimism, including health ben-
e?ts (Scheier & Carver, 1985), improved economic decision mak-
ing (Puri & Robinson, 2007), and task performance (Hales, Wang, &
Williamson, 2014). Future research might, therefore, consider set-
tings in which breaking forecasts or assessments into components
J. Hales / Accounting, Organizations and Society 46 (2015) 19–22 21
could, by generating greater optimism, also lead to performance
improvements. For example, a forecast often becomes a target, as
in CRZ. However, CRZ used a knowledge-based task that might
have left relatively little room for a strong motivational aspect
to performance. In other tasking settings, where performance de-
pends more heavily on effort intensity and/or duration, optimism
in forecasts might translate into improvements in performance.
4. Opportunities for future research more generally
As implied in the introduction, the issue of disaggregation is
fundamental to accounting. Decades ago, a number of theorists
tackled the issue of disaggregation from the opposite perspec-
tive – namely, optimal signal aggregation (e.g., Butterworth, 1972;
Feltham, 1977; Lev, 1968). As Dye and Sridhar (2004) observe, that
line of research was largely c?stymiedd? by Demski (1973), who
applied Blackwell’s (1953) theorem to accounting to argue that
there is no Pareto optimal way to aggregate multiple signals, ab-
sent consideration of how the information will be used. However,
if there are behavioral implications associated with different de-
grees of disaggregation, as shown by CRZ, standard setters, regula-
tors, and control system designers could bene?t from understand-
ing these effects.
Guidance on the issue of disaggregation is important because
it has been a major policy consideration for years, but one that
has resulted in little clear guidance. While U.S. GAAP places few
restrictions on disaggregation in the ?nancial statements, SEC re-
quirements for public companies reduce that ?exibility somewhat,
but still leave management with a large degree of discretion. As a
result, U.S. reporters typically provide a minimal amount of disag-
gregation on the income statement. IFRS requires somewhat more
disaggregation than U.S. GAAP, but still leaves preparers with a fair
amount of discretion.
Experimental research can be particularly insightful on the ef-
fects of aggregation/disaggregation because of their ability to ob-
serve behavior in settings that do not commonly occur in natural
environments. Several experimental studies have recently studied
disaggregation and have, for the most part, documented numer-
ous bene?ts. For example, Hirst, Koonce, and Venkataraman (2007)
use an experiment to show how disaggregation can increase the
credibility of management’s earnings forecasts. In particular, they
provide evidence that the credibility effect of additional disaggre-
gation acts through three channels – by providing a signal about
the precision of management’s information, by making it easier for
users to assess management performance, ex post, and by increas-
ing ?nancial reporting quality.
Consistent with the long-held presumption that c?aggregation
results in a loss of informationd? (Lev, 1968, p. 247), Hewitt (2009)
demonstrates that disaggregation can be useful to users when the
accrual and cash components of earnings have differential persis-
tence, whereas Hales, Venkataraman, and Wilks (2012) show that
disaggregation on the face of the balance sheet can improve lend-
ing decisions. On the control side, Libby and Brown (2013) show
that disaggregation on the face of the income statement results
in lower auditor tolerance for misstatements. However, again, à la
Demski’s production view of accounting information, it is unclear
how disaggregation will affect managers’ reporting decisions. It re-
mains an open question as to how managers will respond to disag-
gregation in equilibrium. Will they anticipate the decreased toler-
ance for misstatement that is likely to come with disaggregation?
As a next step, one could examine how disaggregation changes
manager behavior, even absent anticipated and/or realized changes
in auditor scrutiny.
On this topic, Bonner, Clor-Proell, and Koonce (2014) recently
used a series of experiments to demonstrate that, when given
the opportunity, managers will opportunistically use ?exibility in
aggregating/disaggregating gains and losses from ?nancial instru-
ments consistent with psychological research on mental accounting
(Thaler, 1985) and prospect theory (Kahneman & Tversky, 1979).
Moreover, managers’ preferred presentation results in the highest
valuation from investors, suggesting concordance in manager ex-
pectations of investor perceptions and actual investor perceptions.
Disaggregation is a similarly large and important issue in man-
agement accounting. As CRZ show, it matters for forecasting. It
probably matters for other areas of managerial accounting as well.
During the conference, there was also some discussion about when
optimistic bias might reverse. For example, in budgeting settings
where there are concerns about ratcheting, disaggregation could
lead to increased sandbagging. In budget settings with informa-
tion asymmetry around costs, disaggregated budgets might result
in increased slack, depending on employees’ ability to rationalize
several small rent extractions as opposed to one larger one.
Along these lines, in a recent article, Arya and Glover (2014)
provide a nice theoretical overview of how aggregation can lead to
many such bene?ts. More speci?cally, they discuss how aggrega-
tion could be used to convey information and improve information
quality (e.g., through error cancelation or by limiting cherry pick-
ing in reporting), to boost productivity (e.g., by playing off career
concerns or acting as a commitment device), and to improve the
principal’s ability to control an agent (e.g., by motivating mutual
monitoring).
5. Conclusion
It is my hope that future research will dig deeper into the costs
and bene?ts of disaggregation – both for external ?nancial report-
ing and for internal management budgeting and forecasting. In par-
ticular, the results of CRZ demonstrate that disaggregated forecast-
ing can alter the production of accounting information because of
an interaction with motivated reasoning. This ?nding raises the
question of how disaggregation will interact with other types of
nonmonetary factors, such preferences for honesty, altruism, fair-
ness concerns, and other types of other regarding preferences. Dis-
aggregation could then in?uence not only forecast accuracy, but
also the amount of slack built into budget requests, the tendency
to engage in earnings management, and the accuracy of external
?nancial reports. In my opinion, there is much that remains to be
learned as we once again dig into the most fundamental of ac-
counting questions.
References
Blackwell, D. (1953). Equivalent comparison of experiments. Annals of Mathematical
Statistics, 24, 267–272.
Bloom?eld, R. J. (2002). The c?incomplete revelation hypothesisd? and ?nancial re-
porting. Accounting Horizons, 16(3), 233–243.
Bloom?eld, R. J. (2012). A pragmatic approach to more e?cient corporate disclosure.
Accounting Horizons, 26(2), 357–370.
Bonner, S. E., Clor-Proell, S. M., & Koonce, L. (2014). Mental accounting and disag-
gregation based on the sign and relative magnitude of income statement items.
The Accounting Review, 89(6), 2087–2114.
Butterworth, J. E. (1972). The accounting system as an information function. Journal
of Accounting Research, 10(1), 1–27.
Cain, D. M., Loewenstein, G., & Moore, D. A. (2005). The dirt on coming clean: Per-
verse effects of disclosing con?icts of interest. Journal of Legal Studies, 34(1),
1–25.
Chen, C. X., Rennekamp, K. M., & Zhou, F. H. (2015). The effects of forecast type and
performance-based incentives on the quality of management forecasts. Account-
ing, Organizations and Society.http://dx.doi.org/10.1016/j.aos.2015.03.002.
Christensen, J., & Demski, J. S. (2007). Anticipatory reporting standards. Accounting
Horizons, 21(4), 351–370.
Demski, J. S. (1973). The general impossibility of normative accounting standards.
The Accounting Review, 48(4), 718–723.
Dye, R. A., & Sridhar, S. (2004). Reliability-relevance trade-offs and the e?ciency of
aggregation. Journal of Accounting Research, 42(1), 51–88.
Feltham, G. A. (1977). Cost aggregation: An information economic analysis. Journal
of Accounting Research, 15(1), 42–70.
Hales, J. (2007). Directional preferences, information processing, and investors’ fore-
casts of earnings. Journal of Accounting Research, 45(3), 607–628.
22 J. Hales / Accounting, Organizations and Society 46 (2015) 19–22
Hales, J., Kuang, X., & Venkataraman, S. (2011). Who believes the hype? An experi-
mental investigation of how language affects investor judgments. Journal of Ac-
counting Research, 49(1), 223–255.
Hales, J., Venkataraman, S., & Wilks, T. J. (2012). Accounting for lease renewal op-
tions: The informational effects of unit of account choices. The Accounting Re-
view, 87(1), 173–197.
Hales, J., Wang, L., & Williamson, M. W. (2014). Optimism and contract selection:
Implications for broad-based stock option grants. The Accounting Review. http:
//dx.doi.org/10.2308/accr-50962.
Han, J., & Tan, H.-T. (2010). Investors’ reactions to management earnings guidance:
The joint effect of investment position, news valence, and guidance form. Jour-
nal of Accounting Research, 48(1), 81–104.
Hewitt, M. (2009). Improving investors’ forecast accuracy when operating cash ?ows
and accruals are differentially persistent. The Accounting Review, 84(6), 1913–
1931.
Hirst, D. E., Koonce, L., & Venkataraman, S. (2007). How disaggregation enhances
the credibility of management earnings forecasts. Journal of Accounting Research,
45(4), 811–837.
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under
risk. Econometrica, 47(2), 263–291.
Lev, B. (1968). The aggregation problem in ?nancial statements: An informational
approach. Journal of Accounting Research, 6(2), 247–261.
Libby, R., & Brown, T. (2013). Financial statement disaggregation decisions and au-
ditors’ tolerance for misstatement. The Accounting Review, 88, 641–665.
Arya, A., & Glover, J. C. (2014). On the upsides of aggregation. Journal of Management
Accounting Research, 26(2), 151–166.
Puri, M., & Robinson, D. T. (2007). Optimism and economic choice. Journal of Finan-
cial Economics, 86(1), 71–99.
Scheier, M. F., & Carver, C. S. (1985). Optimism, coping, and health: Assessment and
implications of generalized outcome expectancies. Health Psychology, 4(3), 219–
247.
Thaler, R. (1985). Mental accounting and consumer choice. Marketing Science, 4(3),
199–214.
Thayer, J. (2011). Determinants of investors’ information acquisition: Credibility and
con?rmation. The Accounting Review, 86(1), 1–22.
doc_715348400.pdf
Inthispaper,IprovideasummaryanddiscussionofChen,Rennekamp,andZhou(2015).Doingsoallowsmetheopportunitytodiscussthreeinterrelatedthemes–forecasting,motivatedreasoning,anddisag-gregation.Followingmysummary,Idiscussspecificopportunitiestoextendthislineofresearch.Ialsoofferexamplesofmoregeneralextensionsrelatedtothepsychologyandeconomicsofdisaggregation.
Accounting, Organizations and Society 46 (2015) 19–22
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
Discussion of c?The effects of forecast type and performance-based
incentives on the quality of management forecastsd?
R
Jeffrey Hales
Georgia Institute of Technology, United States
a r t i c l e i n f o
Article history:
Received 22 April 2015
Accepted 23 April 2015
Available online 5 May 2015
a b s t r a c t
In this paper, I provide a summary and discussion of Chen, Rennekamp, and Zhou (2015). Doing so allows
me the opportunity to discuss three interrelated themes – forecasting, motivated reasoning, and disag-
gregation. Following my summary, I discuss speci?c opportunities to extend this line of research. I also
offer examples of more general extensions related to the psychology and economics of disaggregation.
© 2015 Elsevier Ltd. All rights reserved.
1. Introduction
In discussing Chen, Rennekamp, and Zhou (2015, CRZ), I will
touch on three interrelated themes – forecasting, motivated rea-
soning, and disaggregation. To see how these themes relate, we
can start by viewing accounting through the lens of information
economics. In that light, accounting research can be characterized
as the study of how particular types of information (e.g., sales
forecasts, production estimates, external ?nancial reports, etc.) af-
fect the beliefs and actions of particular agents (e.g., managers, in-
vestors, creditors, auditors, etc.). I say affect, but the ?ow of causal-
ity is not unidirectional. As Christensen and Demski (2007) argue,
the information in accounting systems is endogenous because re-
quirements about what and how to measure can alter the supply
of transactions that will be observed. Thus, we cannot take infor-
mation as given, but rather must also understand how information
is created and utilized in equilibrium. Beliefs, therefore, matter and
motivated reasoning has much to say about how information will
be formulated and interpreted.
In a world with strict assumptions about individual rational-
ity and strong form market e?ciency, one could assume away
or marginalize many important accounting questions. However, as
these assumptions are relaxed, the issues of information produc-
tion and consumption become even more interesting as individual
sources of information can no longer be summarily replaced by a
single signal – market price – acting as a su?cient statistic for all
available information. Suddenly, ?nancial statements have meaning
and not just because they are used to calculate earnings. The com-
R
I thank Robert Bloom?eld, Shankar Venkataraman, and Donnie Young for help-
ful comments and Deloitte & Touche LLP for their generous funding of the 2014
Accounting, Organizations & Society Conference on Accounting Estimates.
E-mail address: [email protected]
ponents matter as well.
Within the ?rm, things are similarly complex and interesting.
Production managers make forecasts, superiors review budget re-
quests, and divisions work to meet performance targets. All of
these actions and decisions are based on data and convey infor-
mation. As noted by Butterworth (1972):
nderlying each standard input cost is an implied decision
about the kind and amount of resource to be used and the
nature of the market in which it is to be purchased. Behind
each input/output relationship is an implied decision regarding
the technical conditions under which the process should oper-
ate and the rate at which employees should work. In order to
obtain solutions to one set of problems – those for which the
output of a standard cost system will be considered useful – we
must assume the solution of many others. Often, our assump-
tions turn out to be wrong. The dilemma is universal and is
central to the problem of information system design.
p. 1–2
It is perhaps helpful to draw a distinction between information
and data. Data typically must be analyzed and aggregated in some
fashion to be used. In the words of Bloom?eld (2002), data must
be transformed into statistics to become informative. Nor is this
just a matter of practicality. While it might be infeasible today to
share every bit of business data captured by an enterprise resource
system with investors and creditors, that practical limitation will
someday be gone, if advancements in our ability to capture and
transfer data exceed advancements in our ability to generate data.
While advancements in the world of Big Data make it possible to
capture more and more bits of data, they do not eliminate the need
for aggregation and transformation.
1
1
While I’m referring speci?cally to aggregation and transformation here, I shouldhttp://dx.doi.org/10.1016/j.aos.2015.04.006
0361-3682/© 2015 Elsevier Ltd. All rights reserved.
20 J. Hales / Accounting, Organizations and Society 46 (2015) 19–22
Implicit in this drive to master Big Data is the idea that more
data is better, and often it is – especially when forming statistics.
But it need not always be the case that relying on ?ner parti-
tions of data will result in better statistics and/or improved deci-
sion making. Chen et al. (2015, CRZ) demonstrate one way in which
this could be true, but there are likely others, as I will discuss later.
The format of my discussion is as follows. In the next section,
I provide a summary of the ?ndings in CRZ. Following the sum-
mary, I discuss speci?c opportunities to extend CRZ before closing
with a discussion of more general extensions related to the idea of
disaggregation.
2. Summary of CRZ
CRZ conduct an experiment demonstrating the tradeoffs that
can arise when agents forecast components of a summary mea-
sure of their performance, rather than just directly forecasting the
summary measure itself. As the results of CRZ indicate, forecasting
components results in an improved forecast of the summary mea-
sure (i.e., it is more accurate and no more biased), but only when
agents do not have a monetary incentive linked to the performance
measure. When participants do have a monetary incentive linked
to the performance measure, disaggregated forecasting harms fore-
cast quality, resulting in forecasts that are no longer more accurate
and that are now biased.
Because of the context of their experiment, CRZ ?ts solidly into
the managerial accounting literature on budgeting and forecasting.
Research on management control systems has typically focused on
a single input/output for the control system. CRZ add additional
depth to this literature by noting that, in naturally occurring set-
tings within organizations, essentially every budget request, pro-
duction forecast, or sales estimate will be supported – explicitly
or implicitly – by underlying components of data, reminiscent of
Butterworth (1972). Management control systems could, therefore,
be designed to encourage – or discourage – explicit assessment of
these components.
If all agents were perfectly rational, this particular system de-
sign choice would have no effect on the outcome. But, intuitively,
it is easy to imagine that a control system could act as a decision
aid by encouraging agents to explicitly consider more components
than they might otherwise naturally be inclined to do. In other
words, introducing more decision/assessment points into the de-
cision making process, could combat an important potential short-
coming to optimal decision making – miserly cognition. But, in do-
ing so, it can also open the door to another potential problem –
increased bias in processing.
This is where, from a theoretical perspective, CRZ make another
contribution by extending prior work on motivated reasoning. Prior
accounting research has examined how motivated reasoning alters
investors’ processing of information (Hales, 2007) and subsequent
search (Thayer, 2011). More recently, researchers have also demon-
strated that motivated reasoning in?uences how investors inter-
pret certain types of quantitative management guidance (Han &
Tan, 2010) and certain types of non-quantitative language (Hales,
Kuang, & Venkataraman, 2011). CRZ extend this line of research by
adding what I view as an interesting innovation – examining the
multiple behavioral effects that can arise when a high-level judg-
ment is broken down into components.
In the experiment, CRZ manipulate the incentive schemes their
participants face, which they use to alter the strength of motivated
reasoning. This manipulation has theoretical value, by helping CRZ
also note that the linguistic ?eld of pragmatics has much to say on a related topic –
namely, conveying information through elevation and implicature, as discussed by
Bloom?eld (2012).
to isolate their constructs of interest, and also practical value be-
cause compensation schemes can be altered in practice. Thus, if
?rms need to elicit forecasts from individuals with strong direc-
tional incentives, they may want to encourage relatively high-level
assessments. In contrast, if the forecaster is expected to be rela-
tively objective (i.e., directional incentives tied to the forecast are
relatively neutral), encouraging the forecaster to explicitly assess
lower-level components might be more valuable.
3. Opportunities for extending CRZ
3.1. Understanding the process
CRZ posit that having additional assessments could hurt a bi-
ased forecaster by limiting the bene?t of random error cancelation.
In other words, while disaggregation could help an objective fore-
caster be more accurate (and no more/less biased), having a biased
forecaster explicitly assess components will cause the component
error terms to become correlated. As CRZ argue, the cost of cor-
related error terms could offset the bene?t of getting potentially
smaller error terms, resulting in an aggregate forecast that is more
biased and no more accurate than if the biased forecaster were
to have simply made a high-level forecast. CRZ provide some ev-
idence on this process in their supplemental analysis section, but
a future study could be designed to more carefully look at the un-
derlying process and to examine alternative forecasting techniques
that might result in low error and bias.
For example, instead of forecasting either a high-level fore-
cast or a set of component forecasts, individuals could be encour-
aged/required to do both. In that context, would order matter or
would a c?top-downd? approach and a c?bottom-upd? approach
end up at essentially the same spot? If it is the latter, would that
end result simply be in the middle of the two single step forecast-
ing conditions or would the second stages (having to forecast an
aggregate from components or components from a high-level fore-
cast) be differentially in?uential on the ?nal outcome?
3.2. Task settings
One conference participant raised the topic of multi-product di-
visions. In those settings, bottom-up forecasting might have a dif-
ferent impact as information makes its way up the chain. In par-
ticular, one difference is that, as pieces of information move up the
chain, the strength of the directional preference associated with
that information could weaken because the individuals using the
information as an input are less directly involved in the production
task being forecasted. In other words, we might not see the same
increase in optimism when one person is forecasting the perfor-
mance of other individuals compared to when he or she is break-
ing their own performance up into component pieces. In addition,
as forecasts are moving up the chain, there is always the possibil-
ity that the superior will anticipate some degree of bias and error.
Since bias is directional, it might be easier for a superior to undo
bias in a subordinate’s forecast than to ?lter out unsigned errors.
However, even unraveling anticipated bias can be a tricky prospect
(see, e.g., Cain, Loewenstein, & Moore, 2005).
During the conference, another participant noted that optimistic
bias about future performance is viewed as negative in CRZ, which
seems appropriate given that the task is to convey information.
But one could imagine that optimism might not be all bad. In
fact, research in psychology has documented a number of bene-
?ts associated with dispositional optimism, including health ben-
e?ts (Scheier & Carver, 1985), improved economic decision mak-
ing (Puri & Robinson, 2007), and task performance (Hales, Wang, &
Williamson, 2014). Future research might, therefore, consider set-
tings in which breaking forecasts or assessments into components
J. Hales / Accounting, Organizations and Society 46 (2015) 19–22 21
could, by generating greater optimism, also lead to performance
improvements. For example, a forecast often becomes a target, as
in CRZ. However, CRZ used a knowledge-based task that might
have left relatively little room for a strong motivational aspect
to performance. In other tasking settings, where performance de-
pends more heavily on effort intensity and/or duration, optimism
in forecasts might translate into improvements in performance.
4. Opportunities for future research more generally
As implied in the introduction, the issue of disaggregation is
fundamental to accounting. Decades ago, a number of theorists
tackled the issue of disaggregation from the opposite perspec-
tive – namely, optimal signal aggregation (e.g., Butterworth, 1972;
Feltham, 1977; Lev, 1968). As Dye and Sridhar (2004) observe, that
line of research was largely c?stymiedd? by Demski (1973), who
applied Blackwell’s (1953) theorem to accounting to argue that
there is no Pareto optimal way to aggregate multiple signals, ab-
sent consideration of how the information will be used. However,
if there are behavioral implications associated with different de-
grees of disaggregation, as shown by CRZ, standard setters, regula-
tors, and control system designers could bene?t from understand-
ing these effects.
Guidance on the issue of disaggregation is important because
it has been a major policy consideration for years, but one that
has resulted in little clear guidance. While U.S. GAAP places few
restrictions on disaggregation in the ?nancial statements, SEC re-
quirements for public companies reduce that ?exibility somewhat,
but still leave management with a large degree of discretion. As a
result, U.S. reporters typically provide a minimal amount of disag-
gregation on the income statement. IFRS requires somewhat more
disaggregation than U.S. GAAP, but still leaves preparers with a fair
amount of discretion.
Experimental research can be particularly insightful on the ef-
fects of aggregation/disaggregation because of their ability to ob-
serve behavior in settings that do not commonly occur in natural
environments. Several experimental studies have recently studied
disaggregation and have, for the most part, documented numer-
ous bene?ts. For example, Hirst, Koonce, and Venkataraman (2007)
use an experiment to show how disaggregation can increase the
credibility of management’s earnings forecasts. In particular, they
provide evidence that the credibility effect of additional disaggre-
gation acts through three channels – by providing a signal about
the precision of management’s information, by making it easier for
users to assess management performance, ex post, and by increas-
ing ?nancial reporting quality.
Consistent with the long-held presumption that c?aggregation
results in a loss of informationd? (Lev, 1968, p. 247), Hewitt (2009)
demonstrates that disaggregation can be useful to users when the
accrual and cash components of earnings have differential persis-
tence, whereas Hales, Venkataraman, and Wilks (2012) show that
disaggregation on the face of the balance sheet can improve lend-
ing decisions. On the control side, Libby and Brown (2013) show
that disaggregation on the face of the income statement results
in lower auditor tolerance for misstatements. However, again, à la
Demski’s production view of accounting information, it is unclear
how disaggregation will affect managers’ reporting decisions. It re-
mains an open question as to how managers will respond to disag-
gregation in equilibrium. Will they anticipate the decreased toler-
ance for misstatement that is likely to come with disaggregation?
As a next step, one could examine how disaggregation changes
manager behavior, even absent anticipated and/or realized changes
in auditor scrutiny.
On this topic, Bonner, Clor-Proell, and Koonce (2014) recently
used a series of experiments to demonstrate that, when given
the opportunity, managers will opportunistically use ?exibility in
aggregating/disaggregating gains and losses from ?nancial instru-
ments consistent with psychological research on mental accounting
(Thaler, 1985) and prospect theory (Kahneman & Tversky, 1979).
Moreover, managers’ preferred presentation results in the highest
valuation from investors, suggesting concordance in manager ex-
pectations of investor perceptions and actual investor perceptions.
Disaggregation is a similarly large and important issue in man-
agement accounting. As CRZ show, it matters for forecasting. It
probably matters for other areas of managerial accounting as well.
During the conference, there was also some discussion about when
optimistic bias might reverse. For example, in budgeting settings
where there are concerns about ratcheting, disaggregation could
lead to increased sandbagging. In budget settings with informa-
tion asymmetry around costs, disaggregated budgets might result
in increased slack, depending on employees’ ability to rationalize
several small rent extractions as opposed to one larger one.
Along these lines, in a recent article, Arya and Glover (2014)
provide a nice theoretical overview of how aggregation can lead to
many such bene?ts. More speci?cally, they discuss how aggrega-
tion could be used to convey information and improve information
quality (e.g., through error cancelation or by limiting cherry pick-
ing in reporting), to boost productivity (e.g., by playing off career
concerns or acting as a commitment device), and to improve the
principal’s ability to control an agent (e.g., by motivating mutual
monitoring).
5. Conclusion
It is my hope that future research will dig deeper into the costs
and bene?ts of disaggregation – both for external ?nancial report-
ing and for internal management budgeting and forecasting. In par-
ticular, the results of CRZ demonstrate that disaggregated forecast-
ing can alter the production of accounting information because of
an interaction with motivated reasoning. This ?nding raises the
question of how disaggregation will interact with other types of
nonmonetary factors, such preferences for honesty, altruism, fair-
ness concerns, and other types of other regarding preferences. Dis-
aggregation could then in?uence not only forecast accuracy, but
also the amount of slack built into budget requests, the tendency
to engage in earnings management, and the accuracy of external
?nancial reports. In my opinion, there is much that remains to be
learned as we once again dig into the most fundamental of ac-
counting questions.
References
Blackwell, D. (1953). Equivalent comparison of experiments. Annals of Mathematical
Statistics, 24, 267–272.
Bloom?eld, R. J. (2002). The c?incomplete revelation hypothesisd? and ?nancial re-
porting. Accounting Horizons, 16(3), 233–243.
Bloom?eld, R. J. (2012). A pragmatic approach to more e?cient corporate disclosure.
Accounting Horizons, 26(2), 357–370.
Bonner, S. E., Clor-Proell, S. M., & Koonce, L. (2014). Mental accounting and disag-
gregation based on the sign and relative magnitude of income statement items.
The Accounting Review, 89(6), 2087–2114.
Butterworth, J. E. (1972). The accounting system as an information function. Journal
of Accounting Research, 10(1), 1–27.
Cain, D. M., Loewenstein, G., & Moore, D. A. (2005). The dirt on coming clean: Per-
verse effects of disclosing con?icts of interest. Journal of Legal Studies, 34(1),
1–25.
Chen, C. X., Rennekamp, K. M., & Zhou, F. H. (2015). The effects of forecast type and
performance-based incentives on the quality of management forecasts. Account-
ing, Organizations and Society.http://dx.doi.org/10.1016/j.aos.2015.03.002.
Christensen, J., & Demski, J. S. (2007). Anticipatory reporting standards. Accounting
Horizons, 21(4), 351–370.
Demski, J. S. (1973). The general impossibility of normative accounting standards.
The Accounting Review, 48(4), 718–723.
Dye, R. A., & Sridhar, S. (2004). Reliability-relevance trade-offs and the e?ciency of
aggregation. Journal of Accounting Research, 42(1), 51–88.
Feltham, G. A. (1977). Cost aggregation: An information economic analysis. Journal
of Accounting Research, 15(1), 42–70.
Hales, J. (2007). Directional preferences, information processing, and investors’ fore-
casts of earnings. Journal of Accounting Research, 45(3), 607–628.
22 J. Hales / Accounting, Organizations and Society 46 (2015) 19–22
Hales, J., Kuang, X., & Venkataraman, S. (2011). Who believes the hype? An experi-
mental investigation of how language affects investor judgments. Journal of Ac-
counting Research, 49(1), 223–255.
Hales, J., Venkataraman, S., & Wilks, T. J. (2012). Accounting for lease renewal op-
tions: The informational effects of unit of account choices. The Accounting Re-
view, 87(1), 173–197.
Hales, J., Wang, L., & Williamson, M. W. (2014). Optimism and contract selection:
Implications for broad-based stock option grants. The Accounting Review. http:
//dx.doi.org/10.2308/accr-50962.
Han, J., & Tan, H.-T. (2010). Investors’ reactions to management earnings guidance:
The joint effect of investment position, news valence, and guidance form. Jour-
nal of Accounting Research, 48(1), 81–104.
Hewitt, M. (2009). Improving investors’ forecast accuracy when operating cash ?ows
and accruals are differentially persistent. The Accounting Review, 84(6), 1913–
1931.
Hirst, D. E., Koonce, L., & Venkataraman, S. (2007). How disaggregation enhances
the credibility of management earnings forecasts. Journal of Accounting Research,
45(4), 811–837.
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under
risk. Econometrica, 47(2), 263–291.
Lev, B. (1968). The aggregation problem in ?nancial statements: An informational
approach. Journal of Accounting Research, 6(2), 247–261.
Libby, R., & Brown, T. (2013). Financial statement disaggregation decisions and au-
ditors’ tolerance for misstatement. The Accounting Review, 88, 641–665.
Arya, A., & Glover, J. C. (2014). On the upsides of aggregation. Journal of Management
Accounting Research, 26(2), 151–166.
Puri, M., & Robinson, D. T. (2007). Optimism and economic choice. Journal of Finan-
cial Economics, 86(1), 71–99.
Scheier, M. F., & Carver, C. S. (1985). Optimism, coping, and health: Assessment and
implications of generalized outcome expectancies. Health Psychology, 4(3), 219–
247.
Thaler, R. (1985). Mental accounting and consumer choice. Marketing Science, 4(3),
199–214.
Thayer, J. (2011). Determinants of investors’ information acquisition: Credibility and
con?rmation. The Accounting Review, 86(1), 1–22.
doc_715348400.pdf