This preview shows page 1. Sign up to view the full content.
Unformatted text preview: Yale ICF Working Paper No. 02-37
August 2, 2002 ARE UNMANAGED EARNINGS ALWAYS BETTER
Ohio State University
Jonathan C. Clover
Carnegie Mellon University
Yale School of Management
This paper can be downloaded without charge from the
Social Science Research Network Electronic Paper Collection:
http://ssrn.com/abstract_id=322260 1 Are Unmanaged Earnings Always Better for Shareholders? Anil Arya
Ohio State University Jonathan Glover
Carnegie Mellon University Shyam Sunder
Yale University July 2002 We thank Joel Demski, John Fellingham, Yuji Ijiri, Pierre Liang, Brian Mittendorf, and
Doug Schroeder for helpful comments and discussions. We also thank Patricia Dechow
(special edition editor) and two anonymous referees for helpful comments. 2 Are Unmanaged Earnings Always Better for Shareholders?
SYNOPSIS: The push for increased transparency in financial reporting and corporate
governance serves shareholders only up to a limit. The problem of assessing the value of
transparency to shareholders is subtle because both the level and pattern of earnings can
convey information. Even when earnings management conceals information, it can be
beneficial to shareholders. Distinguishing between ex ante and ex post efficiency
underscores the advantages of achieving a balance between transparency and privacy in
corporations. 3 INTRODUCTION
A lack of transparency in financial reports has often been cited as a weakness of
capital markets. Numerous revelations of financial reporting shenanigans by publicly held
firms have attracted intense political and media attention, of which Enron and WorldCom
are two striking examples. Capital markets are alarmed at such news. The scandals in
2002 have attracted the attention of the securities regulators, accounting standard setters,
Congress, and even the President to corporate earnings management.
These guardians of shareholder interests can be too zealous, even for the good of
their wards. The push for increased transparency in financial reporting and corporate
governance serves the shareholders only up to a point. Beyond that, managerial inhibitions
induced by a lack of privacy can damage the interests of shareholders. In other words,
increasing transparency without limits does not necessarily improve corporate governance.
For extreme example, installing monitoring cameras in offices may inhibit, not motivate
Medical research revealed that cholesterol is not just an artery-hardening villain; it
serves a complex and essential physiological function in our bodies. Similarly, accounting
research shows that income manipulation is not an unmitigated evil; within limits, it
promotes efficient decisions. Our argument, admittedly controversial, is worth airing:
earnings management and managerial discretion are intricately linked to serve multiple
functions; accounting reform that ignores these interconnections could do more harm than
That earnings management reduces transparency is a simplistic idea. A fundamental
feature of decentralized organizations is the dispersal of information across people.
Different people know different things and nobody knows everything. In such an
environment, a managed earnings stream can convey more information than an unmanaged 4 earnings stream. A smooth car ride is not only comfortable; it also reassures the passenger
about the driver's expertise. The remainder of the paper elaborates on these ideas.
A simple story of income manipulation is that the manager owes the best possible
estimate of permanent income of the firm to its shareholders (Fukui 1997). Permanent
income is the periodic income the shareholders can expect in perpetuity; capitalizing this
income stream yields the value of the firm. If permanent income is $10 per year and 12
percent is the appropriate discount rate, the value of the firm is $10/0.12 = $83.33.
Actual income of the firm varies from year to year due to transient shocks as well as
accounting effects. With their access to more information, managers can isolate the
transient from permanent changes better than the shareholders can. If, for example, the
permanent income of the firm remains $10 but a transient factor causes its income in a
given period to be $11, the shareholders, being less well-informed, may capitalize $11 to
arrive at a value of $11/0.12 = $91.67, an error of $8.33 or 10 percent. If, on the other
hand, the increase in income to $11 were permanent, and the less well-informed
shareholders interpreted it to be transient, they would continue to value the firm at $83.33
instead of $91.67; again committing an error of $8.33 in valuation.
Managers could be asked to report permanent changes in income separately from
transient changes to help investors reduce errors in valuation. If transients can be expected
to statistically cancel out over time, managers could report only permanent income which
will be a smoothed out series of raw income numbers. In this scenario, "manipulation" is
not only tolerated, it is encouraged by shareholders for their own interest.
MANAGER WITH PERSONAL GOALS
Suppose, instead of being selfless, managers seek personal goals (e.g.,
compensation). In this case, a manager's desire to manipulate measures of performance 5 becomes more obvious, but why would shareholders want to allow the manager to mislead
them by sending manipulated numbers? This question can be answered in two ways. The
easy answer is: in a world of dispersed information and diverse personal goals of
managers and shareholders, shareholders have no choice. A more subtle and surprising
answer is, even if shareholders had a choice, they may prefer to let managers manipulate
the reports about performance.
The Revelation Principle helps gain perspective on this apparent paradox. The
principle states that if there are no limitations on the three Cs (communication, commitment,
and contract), shareholders and managers can be at least as well off under truthful reporting
as under income manipulation. In other words, breakdown of one or more of the three Cs
is a necessary condition for any degree of income manipulation to be beneficial.
The story in Demski (1998) is one of communicating acquired expertise through
earnings smoothing. The key assumption is that a manager who works hard is both better
able to run the firm, and better able to predict future earnings. The manager demonstrates
his predictive powers, and hence his hard work, to the owner by smoothing earnings.
Because earnings smoothing is an informative variable (the manager is better at it if he
works than if he shirks), smoothing can reduce the cost of motivating the manager to work.
In fact, the owner is better off allowing for earnings smoothing than if she could prevent it
with a perfect and costless audit technology.
An accounting structure in Demski’s model is used to ensure that a manager who
does not work hard finds it difficult to smooth earnings. In particular, Demski assumes
what Sunder (1997) calls the Conservation of Income: the sum of accounting earnings over
the firm’s life is not affected by accounting choices (ignoring the effect of taxes and
changes in the firm's opportunity set). Manipulation catches up with manager. To smooth
earnings well, the manager must be good at forecasting them, and that requires hard work. 6 A feature of Demski’s story is that smoothing is difficult. If the manager can
smooth earnings whether or not he works hard, the owner is always better off contracting
on unmanaged earnings. The implicit role of regulators is to make earnings management
challenging, not impossible. One can view the recent push for increased transparency as an
attempt to make earnings management more challenging.
Demski's view also seems consistent with anecdotal evidence about analysts'
forecasts. A standard story is that when a firm's earnings fall short of the analyst
consensus forecast by even a small amount, there is a large negative stock price reaction
(Skinner and Sloan, forthcoming). This disproportionate reaction might reflect reduced
market confidence in the manager's expertise. The market may assume only an expert
manager can meet the forecast by managing earnings, by managing analysts' expectations,
and/or by performing well.
In Demski’s model, the Revelation Principle is disabled because there is a
restriction on communication--the manager is not allowed to report both current earnings
and a forecast of future earnings. If he could, he would demonstrate his expertise through
his forecast, and unmanaged earnings would be optimal.
Arya and Glover (2001) tell a closely related story. They consider the problem of
selective correction that arises when multiple performance measures are used. Managers
dispute only those performance measures that are understated ("cherry picking"), much as
students dispute the grading of exam questions. The main result in the paper is that cherry
picking can be optimal. This happens when the pre-corrected performance measure is
sufficiently noisy. The idea is similar to Demski’s--cherry picking is itself a variable
informative of the manager’s efforts and, hence, can ease contracting. In contrast to
Demski, the informativeness of performance measure manipulation is endogenous. The
extent to which opportunities for performance manipulation arise depend on the manager's
Limiting Owner Intervention
Distinguishing between ex ante and ex post efficiency can be helpful. For example,
giving students a final exam is ex post inefficient. It imposes the costs of grading time on
the professor, and stress on the students without adding to what they have learned already.
However, from an ex ante perspective, the anticipation of the exam provides the students
with incentives to study hard and learn more. Similarly, earnings management may not be
in the best interest of owners ex post when the earnings report is submitted. However, it
may be in their best interest ex ante when they are trying to induce the manager to join the
firm and exert appropriate effort.
In Arya et al. (1998), earnings management keeps an owner from meddling in
decisions normally left to the manager. For example, when short-term performance is poor
and this is revealed to the owner (or the board of directors), she may take on a greater role
in the day-to-day operations of the firm, participating in decisions normally left to the
manager. The owner would be better off under managed earnings because it keeps her
from intervening excessively in the running of the firm, which is useful ex ante in
motivating the manager. The Revelation Principle breaks down in this case because the
owner’s ability to make binding commitments is limited. A related story is told in Demski
and Frimor (1999). In their story, earnings management limits contract renegotiation.
While the expertise and limited owner intervention stories involve earnings
management, they present very different perspectives. Under the expertise story, the report
distorts current earnings but reveals average earnings and the manager's expertise. Under
the limited owner intervention story, earnings are managed only to conceal information.
This type of earnings management is, arguably, more consistent with the type of
manipulation under attack during 2002. 8
A problem with transparency is that it can encourage a manager to spend his time
signaling his ability (posturing), instead of tending to business. As an example, consider
the following two-period principal-agent model. The key ingredients of the model are: (1)
agents have differing abilities, (2) the principal does not directly observe agents’ abilities,
(3) a hired agent can take action to signal his ability when a transparent reporting system is
installed but not otherwise, and (4) this action may not be in the best interest of the firm.
At the beginning of the first period, the firm's owner (principal) hires a manager
(agent) to run the firm. At the beginning of the second period, the owner either retains the
existing manager or hires a new manager. No long-term commitments are made.
In each period, the firm operates in one of three equally likely environments
(states), s1 , s2 , or s3 . The states are independent across periods. The manager is one of
two equally likely types: rigid or flexible. A rigid manager cannot tailor his act to the
environment the firm faces. In any state, a rigid manager can produce only one output
level, an output of 0. In contrast, a flexible manager can produce two possible output
levels in each state. In s1 , he can produce either -2 or -1; in s2 , -1 or 0; in s3 , 1 or 2.
Assuming the flexible manager produces the highest output he can, the expected output is
1/3, which is more than that of a rigid manager. The manager alone knows his type and the
The output is consumed by the owner at the end of the second period. Employment
yields the manager a private benefit per period that exceeds what he can obtain elsewhere.1
Suppose the information system in place is transparent in that it reveals each
period's output at the end of that period. Assume that the manager, if flexible, acts as the
owner would like him to, i.e., the agent produces the maximum output he can. If the
manager produces 0 in period one, he will be fired. Upon observing 0, the owner's
revised probability (using Bayes Rule) that the existing manager is rigid is 3/4, which is
higher than 1/2, her belief that a new manager will be rigid. 9 Anticipating the owner's firing decision, a flexible manager will posture. That is,
when the state is s2 , the flexible manager produces -1 instead of 0. By producing -1, the
manager will be retained since he has proven he is flexible at the time the firing/retention
decision is made. All that matters to the owner is that the existing flexible manager is better
suited to run the firm in the second period than a new manager--the first period is sunk.
Note the goal incongruence: in state s2 , a flexible manager chooses the wrong act.
The manager is concerned with making himself look good in order to be retained and not
with the firm's actual performance in period one.
Instead, if manipulation were allowed (i.e., manager of either type can claim to
have produced any level of output), the period-one manager would have no credible means
of communicating his type. Hence, the manager cannot use his first period output to
influence the owner's firing/retention decision. This leaves the period-one manager with
the right incentives, which makes the less transparent (manipulable) system optimal.
Under the less transparent system the owner never fires the manager, and the flexible
manager acts as the owner intends–the expected total outcome is 1/3. Under the more
transparent system, the owner fires the manager if and only if the period one outcome is 0,
and the flexible manager acts as the owner intends, except when the state is s2 --the
expected total outcome is 1/4.
In our example, the problem with transparency is it provides the manager with
incentives to choose an action that demonstrates his versatility even when such action is
detrimental to the firm. By insisting on less transparent reporting, the owner can assure the
manager he can "do the right thing" without the fear of being mislabeled.2 More generally
earnings management is one of many ways of aligning incentives, and the optimal mix of
instruments may include giving the manager discretion in reporting.
Although perfect transparency (observing the manager's type as well as all other
information) is best for the owner, no information is preferred over some information.3
Also, allowing for reporting manipulation can be a useful way of discouraging real 10 manipulation (choosing the wrong act). When we change accounting we need to be careful
to anticipate associated real effects (for example, underinvestment in research and
development when the manager is forced to expense all such costs).
In the posturing story, the Revelation Principle breaks down because the owner’s
ability to make binding commitments is limited. This leaves contract restrictions as the only
assumption of the Revelation Principle we have not discussed in this paper. In Liang’s
(2001) recent work, earnings management is a means of introducing non linearity into an
otherwise linear contract.
Researchers have devoted much effort in trying to document earnings management in
empirical studies. The evidence to date remains weak and inconsistent (Dechow and Skinner
2000; Sunder 1997). Reasons include the inherent endogeneity and unobservability of
earnings management, and the focus in most studies on a single instrument of earnings
management. Joel Demski wrote in Accounting Education News (Spring 2002): "Consider
the broad topic of earnings management. Drawing the circle narrowly around the subject, we
tend to proceed with the presumption such behavior exists and focus on its documentation.
Yet a broader approach would treat the behavior as endogenous and focus on the web of
institutional and organizational arrangements that precede any ability to detect that behavior."
In his February 26, 2002 testimony before the US Senate, Walter P. Schuetze (a
former Chief Accountant of the SEC) stated “Earnings management is a scourge in this
country …We need to put a stop to earnings management.” He went on to argue for markto-market accounting in order to provide a true picture of firms. Equating the relationship
between financial reporting and business to that of a photographer and landscape is
inherently problematical. It is more like that of a photographer to a model: the model smiles
and poses at the camera even as the photographer changes camera angle and settings in 11 reaction to the model. The state of the firm and its financial reports are reflexive in the
sense of being dependent on each other. Debates on financial accounting reforms often
focus on making changes to deal with current problems, failing to anticipate problems that
may emerge as agents respond strategically to the reforms.
Instead of trying to eliminate earnings management, it might be more useful to
emphasize accounting properties that increase the value of managed accruals. We briefly
discuss three such properties. First, an often-discussed feature of accounting is its
verifiability. (See Ijiri  on the related notion of hardness.) Accounting complements
other less verifiable information sources by its ability to discipline them. Arguably, as
accounting has become more and more inclusive and focused on market values, its
comparative advantage has eroded.
Second, accounting recognizes only select events and the choice of accounting
methods effectively locks managers into the future treatment of these select events (see
Pacharn 2002). Third, the Law of Conservation of Income ensures accounting earnings
are correct in total over the life of the firm. One might take this a step further and argue that
a desirable feature of US GAAP is (or should be) that things catch up relatively quickly.
Suppose a firm has a bad year, and the CEO adopts aggressive accounting to cover up the
bad news in the short run with the expectation that he or she will turn things around next
year and be able to make up for the shortfall of the previous year. If instead the next year is
also a bad year, things may blow up on the manager in the sense that an even larger
overstatement of earnings is needed to maintain the earnings growth the market may have
come to expect.
These three properties are examples of the curious blend of rigidity and flexibility
that accounting exhibits. Accruals are flexible, yet verifiability, selective recognition, and
conservation enforce certain tidiness and keep them from straying too far (see Demski et al.
). A natural focus for policy debates is the optimal mix of rigidity and flexibility,
which is unlikely to be a corner solution. 12 REFERENCES
Arya, A. and J. Glover. 2001. Aggregate performance measures as a response to one-sided
error correction. Working Paper, Carnegie Mellon University.
Arya, A., J. Glover, and S. Sunder. 1998. Earnings management and the Revelation
Principle. Review of Accounting Studies 3: 7-34.
Cremer, J. 1995. Arm's length relationships. Quarterly Journal of Economics 110: 275-295.
Dechow, P. and D. Skinner. 2000. Earnings management: Reconciling views of accounting
academic, practioners, and regulators. Accounting Horizons 14: 235-250.
Demski, J. 1998. Performance measure manipulation. Contemporary Accounting Research
Demski, J. and H. Frimor. 1999. Performance measure garbling under renegotiation in
multi-period agencies. Journal of Accounting Research 37: 187-215.
Demski, J., John Fellingham, Jonathan Glover, Yuji Ijiri, Pierre Liang, and Shyam Sunder.
2002. Some Thoughts on The Intellectual Foundations of Accounting. Accounting
Horizons 16:2, 157-168.
Fukui, Y. 1996. Earnings management: not for managers but for investors. Working paper,
Carnegie Mellon University.
Ijiri, Y. 1975. Theory of Accounting Measurement. Sarasota, FL: AAA.
Liang, P. 2001. The equilibrium reliability of accounting measures. Working paper, Carnegie
Pacharn, P. 2002. Decentralized Accounting Choice and Managerial Incentives. Dissertation,
Carnegie Mellon University.
Skinner, D. and R. Sloan. 2002. Earnings surprises, growth expectations and stock returns.
Forthcoming in Review of Accounting Studies.
Sunder, S. 1997. Theory of Accounting and Control. Cincinnati, OH: South-Western
Publishing. 13 Endnotes 1. We ignore compensation. Compensation can be incorporated without changing the
result as long as the private benefit of employment to the manager is large enough. 2. Our story is closely related to Cremer (1995) in which a principal sometimes finds it
optimal to install a coarse information system in order for her to make it self-enforcing
for her to fire the manager when poor performance is observed (but the manager is
talented). The distinguishing feature of our story is the manager's incentive to posture
(signal his type). 3. Under perfect transparency, the owner fires the rigid manager and retains the flexible
manager, and the manager acts as the owner intends–the total expected outcome is 5/12. ...
View Full Document
This note was uploaded on 02/26/2012 for the course ECON 150 taught by Professor Akerlof during the Spring '12 term at Syracuse.
- Spring '12