Description
The purpose of this paper is to explore the role of accounting discretion in a
principal-agent setting, wherein accounting information is used in performance evaluation. The agent
may choose one from among several allowable accounting methods. However, limited audit resources
allow only for verification of only the method the agent chooses, and this is the only one used to
determine the agent’s compensation.
Accounting Research Journal
A few stylized observations on accounting discretion
Steven Schwartz Richard Young
Article information:
To cite this document:
Steven Schwartz Richard Young , (2013),"A few stylized observations on accounting discretion",
Accounting Research J ournal, Vol. 26 Iss 2 pp. 154 - 166
Permanent link to this document:http://dx.doi.org/10.1108/ARJ -J ul-2012-0064
Downloaded on: 24 January 2016, At: 21:17 (PT)
References: this document contains references to 34 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 482 times since 2013*
Users who downloaded this article also downloaded:
Amanda J . Carter, Roger L. Burritt, J ohn D. Pisaniello, (2013),"The dual roles for accountants in sustaining
rural communities", Accounting Research J ournal, Vol. 26 Iss 2 pp. 130-153http://dx.doi.org/10.1108/
ARJ -12-2012-0093
Victoria J . Clout, Larelle Chapple, Nilan Gandhi, (2013),"The impact of auditor independence regulations
on established and emerging firms", Accounting Research J ournal, Vol. 26 Iss 2 pp. 88-108 http://
dx.doi.org/10.1108/ARJ -Dec-2011-0045
Vijaya Murthy, J ames Guthrie, (2013),"Accounting for workplace flexibility: Internal communication
in an Australian financial institution", Accounting Research J ournal, Vol. 26 Iss 2 pp. 109-129 http://
dx.doi.org/10.1108/ARJ -12-2012-0096
Access to this document was granted through an Emerald subscription provided by emerald-srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.
*Related content and download information correct at time of download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
A few stylized observations
on accounting discretion
Steven Schwartz
School of Management, Binghamton University, Binghamton,
New York, USA, and
Richard Young
Fisher College of Business, The Ohio State University,
Columbus, Ohio, USA
Abstract
Purpose – The purpose of this paper is to explore the role of accounting discretion in a
principal-agent setting, wherein accounting information is used in performance evaluation. The agent
may choose one from among several allowable accounting methods. However, limited audit resources
allow only for veri?cation of only the method the agent chooses, and this is the only one used to
determine the agent’s compensation.
Design/methodology/approach – This paper uses the model developed by Penno (1990) and
construct relatively simple examples that illustrate the effects of accounting discretion on performance
evaluation. These examples may make the implications of Penno (1990) clearer to educators and
practitioners.
Findings – Accounting discretion can be useful in performance evaluation even if the accounting
method chosen is not disclosed. Lower limits on compensation reduce the usefulness of accounting
discretion. If the principal declares a preferred accounting treatment, it enhances the usefulness of
accounting discretion; in fact, it guarantees that accounting discretion is weakly preferred to no discretion.
Practical implications – Accounting discretion is often viewed pejoratively. Those responsible for
designing performance evaluation schemes should be aware of the potential bene?ts of accounting
discretion, and those charged with promulgating accounting rules should consider the possibility that
rules can be too in?exible.
Originality/value – This paper provides accessible illustrations of the effects of accounting
discretion on performance evaluation and offer discussion to place the conclusions drawn from the
model into a standard setting context.
Keywords Performance evaluation, Agency theory, Accounting discretion, Performance appraisal,
Accountancy
Paper type Research paper
Managers clearly have discretion in the way they report their ?rms’ ?nancial results.
Such discretion is often portrayed in the popular press as undesirable (Briloff, 1972;
Levitt, 1998). In many cases, this view is justi?ed. For example, managers at Enron took
liberal advantage of the rules regarding mark-to-market accounting and the
consolidation of special purpose entities (Schwarcz, 2002), while managers at waste
management exploited accounting rules regarding changes in the useful life of
depreciable assets. With the bene?t of hindsight, it is clear both ?rms went beyond the
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1030-9616.htm
The authors would like to thank participants at the Northeast Regional Meeting of the American
Accounting Association and The Ohio State University Accounting Honors students for their
helpful comments and suggestions.
Accounting Research Journal
Vol. 26 No. 2, 2013
pp. 154-166
qEmerald Group Publishing Limited
1030-9616
DOI 10.1108/ARJ-Jul-2012-0064
ARJ
26,2
154
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
limits allowed by generally accepted accounting principles (GAAP). Yet there was
enough discretion allowed in GAAP for each ?rms’ ?nancial reports to receive a clean
bill of health from their auditor, Arthur Andersen, despite the auditor’s general
knowledge of what the ?rm had done (Benston, 2006; SEC, 2001; Tof?er and Reingold,
2003). The academic literature also tends to view accounting discretion pejoratively,
mainly as a means by which managers can expropriate wealth from the ?rm. Examples
include managers using accounting discretion so as to maximize the value of their
holdings of equity options or to ward off takeover attempts that might otherwise bene?t
shareholders (Bergstresser andPhillipon, 2006; Christie and Zimmerman, 1994; Salamon
and Smith, 1979).
We describe and illustrate an alternative view of discretion, using the stylized model
of Penno (1990). We demonstrate that when a principal, such as a board of directors,
grants accounting discretion to an agent, such as a CEO, it may decrease agency costs.
Underlying our analysis is an assumption that sophisticated boards of directors, when
devising incentive contracts, anticipate how managers will use their discretion
(Bowen et al., 2008; Watts and Zimmerman, 1990). Given that boards do so, optimal
contracts that allow for accounting discretion can be bene?cial. Consideration of this
possibility is important not only from the perspective of contract design, but also
policy decisions on the ?exibility of accounting standards.
Penno (1990) models accounting discretion as the ability of the manager to choose
one signal from a vector of signals with common knowledge probability distribution.
The signal chosen cannot be misrepresented (the so-called “anti-fraud rule”), but only
the manager knows the value of the other signals. Penno shows that allowing an agent
a choice of which signal to report provides contracting bene?ts to the principal, despite
the fact that the agent will choose the signal that puts him in the best light. Ozbilgin
and Penno (2008) explore the conditions under which the owner wishes to delegate the
choice of signal to the manager or retain the decision rights to choose the signal[1].
We use the Penno (1990) framework to illustrate four stylized observations about
accounting discretion in a performance evaluation setting. First, accounting discretion
may be bene?cial even if a manager discloses neither his choice of the accounting
methodnor the outcome of the accounting methods he did not use. Second, a lower bound
on a manager’s compensation decreases the effectiveness of accounting discretion in
performance evaluation. Third, if a manager is required to disclose his chosen
accounting method and there is no lower limit on compensation, accounting discretion is
strictly preferred. Fourth, if a manager is required to disclose his chosen accounting
method, a lower bound on compensation may diminish the bene?t of accounting
discretion, but nevertheless more discretion is still weakly preferred to less.
Unlike the literature on accounting disclosure and equity pricing, from a contracting
perspective accounting discretion can be seen as useful under a robust set of
circumstances. In addition, the arguments we make herein do not rely on the notion
that accountants need discretion in order to better communicate the value of the ?rm
(Chambers et al., 2003). In fact, despite what seems to be implied by the language of
accounting, interpreting accounting numbers as economic value and income can be
troublesome in a world where markets are not frictionless. “Therefore, it is important
to look at the usefulness of accounting information as being something other than an
input in the estimation of ?rm value” (See, for example, Christensen and Demski, 2003;
Kanodia et al., 2005)[2]. We end with a brief look at one form of accounting based
Observations
on accounting
discretion
155
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
contracts, debt covenants. We note that creditors often allow debtors accounting
discretion although contracts could be written otherwise.
Model
No reporting discretion (one signal)
Consider a simple binary action agency model, as in Antle and Demski (1988). A ?rm
consists of an owner (principal) and a manager (agent) who contract over a single task.
The manager chooses an action a, where a [{H, L}. The manager has a private cost of
H but no private cost for choosing L: c
H
. c
L
¼ 0. We assume action H is suf?ciently
bene?cial to the owner that he wishes to motivate H, even though he will have to pay
the manager a greater amount for H than L.
The action chosen by the manager is not observable to the owner and so cannot be
contracted upon directly. However, the owner and manager can contract upon a signal
that the manager can in?uence. The signal may be interpreted as the product of the
accounting system. The manager’s report of the signal is veri?ed without error by an
auditor. Assuming a suf?cient penalty is available for a report that is inconsistent with
the auditor’s ?ndings, the manager will in equilibrium report the signal truthfully. For
simplicity, assume the signal has a binary outcome that is either g (good) or b (bad). Let
p be the probability of g given a ¼ H and q be the probability of g given a ¼ L, where
p . q, justifying the labels of “good” and “bad” for g and b, respectively.
The manager is risk-averse and has a utility for money equal to
????
w
p
. In the
contracting program the risk-neutral owner minimizes the expected wage paid to
the manager subject to two constraints. The “participation” constraint (P) guarantees
the manager an expected utility of at least
U, which is the expected utility obtained if
he went elsewhere. The “incentive compatibility” constraint (IC) ensures the choice of H
is weakly better for the manager than L.
The risk neutral owner’s problem to ef?ciently motivate H is found below:
min
w
g
;w
b
pw
g
þ ð1 2pÞw
b
subject to:
p
??????
w
g
p
þ ð1 2pÞ
?????
w
b
p
2 c
H
$
U ðPÞ
p
??????
w
g
p
þ ð1 2pÞ
?????
w
b
p
2 c
H
$ q
??????
w
g
p
þ ð1 2qÞ
?????
w
b
p
ðICÞ
In illustration 1, we set p ¼ 0.7, q ¼ 0.5, c
H
¼ 50 and
U ¼ 500. The solution, found in
Table I, is w
g
¼ 390, 625, w
b
¼ 140,625 and E[w] ¼ 315,625. We can think of the
contract as a base wage of (c
H
þ U)
2
¼ 302,500, which compensates the manager
for the disutility of H and for choosing to participate in the ?rm, with a bonus of 88,125
for a good report and a penalty of 161,875 for a bad report. The bonus and penalty are
necessary to motivate the desired action by the manager when the action is not directly
observable. The expected wage of 315,625 is greater than base wage of 302,500 which
re?ects the need to compensate the risk-averse manager for receiving an uncertain
wage. An improved information system would reduce the amount of risk placed on the
manager; this is what we search for when we introduce accounting discretion.
ARJ
26,2
156
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Accounting discretion
In the previous sub-section the accounting system produced only one signal, to model
the idea of no discretion. Now assume the accounting system generates two
independently distributed signals. The different signals canbe thought of as the result of
different allowable accounting methods, for example FIFO and LIFO. The manager, at
his own discretion, submits a report that must be equal to one of the two signals he
privately observes. The chosen signal cannot be misreported, but perhaps due to limited
auditing resources, only the chosen signal will be veri?ed and reported to the owner.
In this section the agent cannot credibly disclose which of the two signals he chose,
so upon a report of g, the owner only knows that at least one of the two signals is g.
With these assumptions g continues to be good news about the manager’s action
choice. Therefore, the manager will choose to report g, whenever possible, implying a
report of b indicates both signals are b. We maintain the assumptions of p ¼ 0.7 and
q ¼ 0.5, so that the probability of being capable of reporting g is equal to
1 2 (1 2 p)
2
¼ 0.91 conditional on H and 1 2 (1 2 q)
2
¼ 0.75 conditional on L[3]. The
owner’s program for this and subsequent illustrations is found in the Appendix, and
the solution to the program is shown in Table I as illustration 2.
From Table I we observe the cost of motivating H has decreased from 315,625 to
310,498 with the introduction of accounting discretion. This is one of the primary
insights in Penno (1990) – accounting discretion can be valuable from a contracting
perspective. The key to understanding this result is to examine the likelihood ratios that
correspond to the two feasible reports (Antle and Demski, 1988). The likelihood ratio is
the probability of a particular report conditional on L divided by the probability of the
same report conditional on H. A likelihood ratio greater than 1 is “evidence” that L was
taken; a likelihood ratio less than 1 is evidence H was taken[4]. A likelihood ratio of
exactly 1 would imply that observing that particular report is not helpful in determining
which action was taken.
When discretion is added in this example the likelihood ratio for a report of
g moves closer to 1 (from 0.714 to 0.824) and the bonus decreases to
(334,228.5 2 302,500) ¼ 31,728.5. As the likelihood ratio moves closer to 1 the evidence
that H was taken becomes weaker, so it is ef?cient for the owner to reduce the bonus.
Intuitively, increasing the number of potential signals increases the probability g will
be reported regardless of the action taken, hence the bonus decreases.
Illustration 1 Illustration 2 Illustration 3 Illustration 4
Number of signals 1 2 1 2
Conditional
distribution p ¼ 0.7; q ¼ 0.5 p ¼ 0.7; q ¼ 0.5 p ¼ 0.97; q ¼ 0.96 p ¼ 0.97; q ¼ 0.96
Parameters c
H
¼ 50;
U ¼ 500 c
H
¼ 50;
U ¼ 500 c
H
¼ 1;
U ¼ 2,500 c
H
¼ 1;
U ¼ 2,500
Pr(report ¼ gjL) 0.5 0.75 0.96 0.9984
Pr(report ¼ gjH) 0.7 0.91 0.97 0.9991
LR(g) 0.714 0.824 0.9897 0.9993
LR(b) 1.667 2.78 1.3333 1.7778
w
g
390,625 334,228.5 6,270,016 6,261,433
w
b
140,625 70,556.4 5,779,216 1,152,862
E[Wage] 315,625 310,498 6,255,292 6,256,836
Var[LR] 0.190 0.313 0.0034 0.0005
Table I.
Results for illustrations
1, 2, 3 and 4
Observations
on accounting
discretion
157
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Conversely, inmovingfromone signal to two signals the penalty for a report of b increases
to (302,500 2 70,556.4) ¼ 231,943.6, because the likelihood ratio moves further from
1 (from 1.67 to 2.78). Intuitively, increasing the number of potential signals decreases the
likelihood that b will be reported regardless of the action taken, hence the penalty for a
report of b increases.
Ideally, we would like both likelihood ratios to be as far from their expected value
(equal to 1) as possible. Therefore, it is perhaps not surprising that the greater the
“spread” of the likelihood ratios, the more valuable the information systemin evaluating
the manager. Two likelihood ratios of 1 would indicate a system with no performance
evaluation information. Two likelihood ratios that approach 0 and in?nity, for the good
and bad signal, respectively, would indicate a system that provides near-perfect
information. More formally, with a square root utility function for the manager the value
of the accounting system is strictly increasing in the variance of the likelihood ratios
(Kim and Suh, 1991). For the case of no discretion, the variance of the likelihood ratios is
0.7(0.714 2 1)
2
þ 0.3(1.667 2 1)
2
¼ 0.190[5]. For the case of discretion, the likelihood
ratio is 0.313 . 0.190. Comparing the expected wages in illustrations 1 and 2, the owner
prefers to allow discretion on the part of the manager.
So far we have shown that accounting discretion may be valuable even if the owner is
unaware of the accounting method chosen. However, this result is not general. Let us
alter the conditional probability of the signals so that p ¼ 0.97 and q ¼ 0.96. Further let
us assume that c
H
¼ 1 and
U ¼ 2,500. With these new parameters, the no discretion
case is shown in Table I as illustration 3 and the two signals case is shown in Table I as
illustration 4. Our reason for presenting illustrations 3 and 4 is immediate: whether
accounting discretion is bene?cial depends on the details. With the parameters in
illustrations 3 and 4, expected wages increase when discretion is added. As mentioned,
there are countervailing forces at work when accounting discretion is introduced.
A signal of b becomes more informative while a signal g becomes less informative. In
illustration 4 the loss of information fromg outweighs the gain of information fromb[6].
The setting we have examined so far is not uncommon in ?nancial reporting.
Although ?nancial statements contain considerable documentation on accounting
method choice, clearly not all choices are transparent. Yet, we rely on auditors for
assurances that all choices ultimately fall within the realm of GAAP. That is, we know
(or hope) that the report comes from an allowable method but we may not be aware of
exactly the method chosen. The fact that accounting discretion can still be valuable
under these assumptions is both surprising and important in explaining the wide
discretion available in ?nancial reporting.
Bankruptcy constraints
In almost all realistic settings involving performance evaluation and compensation
there is a lower bound on employee remuneration. One useful representation of this
lower bound is a bankruptcy constraint such that employees must receive a
non-negative wage. This makes sense because employees may be ?red but typically do
not pay their ?rm a penalty that exceeds their wages[7]. We explore the effect of a
bankruptcy constraint with illustrations 5 and 6.
In illustration 5, we assume a single signal with the conditional distributions
used in illustrations 1 and 2: p ¼ 0.7 and q ¼ 0.5. However, we change
U to
50 while maintaining c
H
¼ 50. This has the effect of making a bankruptcy constraint
ARJ
26,2
158
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
on w
b
binding. The results are shown in Table II, where we observe the manager
receives 62,500 for g and 0 for b. In comparison, the ?rst-best solution (no information
asymmetry) would pay (c
H
þ U)
2
¼ 10,000, so we may interpret the contract as paying
the manager a bonus of 52,500 for a good outcome and a penalty of 10,000 for a bad
outcome. Recall, when
U ¼ 500 the penalty was almost twice the size of the bonus
(161,875 compared to 88,125). As one might surmise, the owner would like to implement
a similar ratio of penalty to bonus in illustration 5, but cannot due to the bankruptcy
constraint. It should come as no surprise then that as we increase the informativeness of
b and decrease the informativeness of g through accounting discretion, satisfying the
bankruptcy constraint will become increasingly costly and the owner will be made
worse off with discretion. As an analogy, imagine the criminal justice system worked
only be giving those citizens who have been “proven” to be innocent a bonus, with the
further constraint that most people would appear innocent regardless of whether they
had committed a crime. Such a system would be both costly and inef?cient.
In illustration 6 the manager has discretion over two signals with the same
parameters as in illustration 5. The payment for a good report is now97,656.25 while the
payment for a bad report remains at 0. Relative to the no discretion case, expected wages
increase from43,750 to 88,867. The wage for a bad report was already at the lower bound
of zero in illustration 5. Adding a second signal in illustration 6 causes the good outcome
to become even less informative, so the ef?cient contract must increase the bonus even
further (with no ability to increase the penalty) and expected compensation increases.
The overarching story here is that in a performance evaluation setting a lower bound on
compensation can make accounting discretion particularly unfavorable for the owner.
Two observations are in order. As noted, in almost all practical situations
compensation is bounded from below. Most employees cannot be forced to pay back
money to the ?rm except in cases of theft, so that dismissal would be the strongest
penalty. For CEOs and other top executives, even dismissal may be very pecunious
(Bebchuk and Fried, 2003). Therefore, the usefulness of discretion may seem to be
limited, due to restrictions on penalties that boards of directors can realistically impose.
Second, Ozbilgin and Penno (2008) demonstrate that when there is a lower bound on
compensation discretion may still be useful, but only in the case where it is optimal to
give discretion to the owner. With the owner retaining discretion over which signal is
chosen, the bankruptcy constraints for the manager would not be problematic. The
owner would choose the signal that corresponds to the lowest wage, if possible. Hence,
report of a g implies all signals are g and further implies g becomes the more
informative signal. In consequence a report of g is accompanied with a large bonus
while a report of b is accompanied by a small penalty, which is helpful when the
bankruptcy constraint is binding.
Illustration 5: one signal Illustration 6: two signals
Pr(gjL) 0.5 0.75
Pr(gjH) 0.7 0.91
w
g
62,500 97,656.25
w
b
0 0
E[Wage] 43,750 88,867.19
Table II.
Likelihood ratios (LR) for
illustrations 5 and 6
p ¼ 0.7; q ¼ 0.5;
c
H
¼ 50;
U ¼ 50
Observations
on accounting
discretion
159
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Accounting preferences and customs
To this point we have assumed the signals are indistinguishable. We now assume the
manager can credibly disclose which of the signals is being reported. This is in essence
what occurs when the manager provides details in the footnotes about the particular
accounting method chosen. We maintain the assumption that the manager only reports
one signal, and does so in his own best interest.
Let us ?rst return to the parameters used in illustrations 3 and 4. We demonstrated
earlier the owner would be better off if she only allowed the manager only a single
signal. Suppose now that the owner establishes a preference for the manager to report
from one of the two signals. While it is not valuable to force the manager to divulge
which of the two signals he is reporting per se, it is interesting that it is valuable if the
owner establishes a preference ex ante. The reason is simple: if the non-preferred signal
is disclosed as g it can be inferred that the preferred signal is b. Clearly, the owner
potentially receives an additional piece of information if he announces a preference for
the use of a particular signal, relative to the case of no preference. We demonstrate with
illustration 7, found in Table III.
In illustration 7, we label one signal the preferred signal; the other the non-preferred
signal. The probability the manager reports g from the preferred signal conditional on
action H is 0.97, while the probability that the manager reports g from the non-preferred
system conditional on action H is (1 2 0.97)0.03 ¼ 0.0291. Finally, the probability the
manager cannot report g conditional on His equal to 0.03(0.03) ¼ 0.0009. The solution to
illustration 7 is provided in Table III, where w
pg
and w
ng
are the wages for a report of g
from the preferred and non-preferred signals, respectively. The program to solve for the
optimal contract for this illustration (and illustration 8) is in Appendix 2. The expected
wage payment has decreased from 6,255,292 with a single signal (no discretion) to
6,255,277 with two signals and granting discretion to the manager.
Illustration 7 re?ects the general result that as long as the owner announces a
preference for a particular preference, and with no binding bankruptcy constraint, the
owner is strictly better off increasing discretion by adding a new system, as long as the
new system is informative given the ?rst[8]. The intuition is as follows. If the owner
sets a preference, she will always know if the signal coming from the preferred signal is
g or b, as in the case of a single signal. However, she will also know whether the
non-preferred signal is g or b, given the preferred signal is b. Technically speaking,
there is an additional partition of the information space. Hence, adding an additional
Illustration 7: p ¼ 0.97, q ¼ 0.96,
c
H
¼ 1,
U ¼ 2,500
Illustration 8: p ¼ 0.7, q ¼ 0.5,
c
H
¼ 50,
U ¼ 50
Pr( pgjL) 0.96 0.5
Pr( pgjH) 0.97 0.7
Pr(ngjL) 0.0384 0.25
Pr(ngjH) 0.0291 0.21
w
pg
6,269,248 62,500
w
ng
5,821,566 0
w
b
5,221,210 0
E[Wage] 6,255,277 43,750
Table III.
Illustrations 7 and 8 –
accounting discretion
with a preferred signal
ARJ
26,2
160
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
system cannot be harmful to the owner, and without a binding bankruptcy constraint it
is strictly bene?cial. Note the very robust result here.
In illustration 8 we continue to assume that the owner has a preference, but we return
to the parameters from illustrations 5 and 6. Recall, in illustrations 5 and 6 accounting
discretion was harmful to the owner due to the lower bound on compensation to the
manager. Under an optimal contract, due to the binding bankruptcy constraint, w
pg
is
the only non-zero wage. The expected wage of 43,750 is equal to that resulting from a
single signal. In essence the contract replicates the no-discretion contract, the owner does
not gain, but does not lose either. As with illustration 7, the preference causes a partition
in the information space, but in this instance there is no bene?t to using it[9]. This
example illustrates the general result that as long as a preference is established up front
by the owner, there is no information lost adding a new signal, and hence even with a
bankruptcy constraint, accounting discretion is weakly preferred.
It is interesting that boards of directors do not typically state preferred accounting
methods in their executive compensation schemes. Also, market based compensation,
such as employee stock options, has become the dominant form of executive
compensation ( Jensen et al., 2004). This perhaps is where the value of accounting
conventions or customs comes into play. A custom can be thought of as a habitual
practice or behavior that people engage in and expect others to engage in. Customs can
have social bene?t. For example, farmers’ markets being held after certain holidays
have a coordination bene?t. Similarly, when discussing accounting conventions the
American Accounting Association’s Financial Accounting Standards Committee noted
“[accounting conventions] may be valued indirectly for their coordination bene?ts that
even arbitrary choices yield through their social acceptance and wide use”
(American Accounting Association, 2011, p. 11). The market might understand
which accounting method is conventional, and so would react negatively if a method
that is technically acceptable, but not conventional, were used, This phenomenon is
similar to the idea of a preference in our model, in that the market is rightly inferring a
worse report would have resulted had the conventional method been used. Ultimately,
depending on the circumstances, allowing the manager discretion may be ex ante
ef?cient from a contracting perspective.
Discussion and conclusion
Accounting is fundamentally a source of information (Christensen and Demski, 2003).
Accounting information may be useful in valuing a ?rm, determining a ?rms’ tax
liability or evaluating whether a ?rm has met regulatory requirements. Therefore, in
selecting an accounting information regime one should be mindful of its purpose.
In this paper our focus is on accounting’s role in providing useful information for
performance evaluation. We illustrate that accounting discretion can be useful in this
endeavor. One natural question that arises though with respect to our analyses is: do
contract designers, such as boards of directors, consider the effects of accounting
discretion when designing contracts? This may seem like lot to ask from Board
members. However, Coles et al. (2006) provide evidence that market participants
anticipate earnings management around the re-issuance of stock options, so it may not
be so unreasonable that boards of directors anticipate the behavior as well. The situation
we model is similar to accounting discretion and accounting-based debt covenants. We
know a lot about debt covenants, so it might be fruitful to look there for contracting
Observations
on accounting
discretion
161
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
evidence related to accounting discretion. Creditors can easily write contracts restricting
discretion over accounting method and indeed charge lower interest rates when doing so
(Beatty et al., 2002). However, creditors generally allow discretion, and research has
shown that debtors use the discretion to avoid default (Beatty and Weber, 2003). One
possibility is that for some creditor-debtor relationships accounting discretion is part of
an ef?cient contract.
Of more direct relevance is whether accounting based performance evaluation
schemes consider the accounting choices of managers[10]. There is some evidence that
managers are evaluated differently based on how they report income. For example,
non-recurring charges and losses due to forced changes in accounting procedure tend to
not affect the manager’s compensation (Dechow et al., 1994;Kren and Leauby, 2001).
However, these are really measures to protect the manager against circumstances
beyond his control, rather thanto encourage the use of a particular accounting treatment.
It has been pointed out that it may be costly to proscribe certain accounting treatments,
because managers have a better understanding of which accounting methods are most
bene?cial for the ?rm from an overall perspective (e.g. in aiding make-or-buy decisions
and pricing decisions) rather than from just a performance evaluation perspective
(Christie and Zimmerman, 1994). In fact, Bowen et al. (2008) provide evidence that
greater accounting discretion may be associated with better ?rm performance in the
future. Therefore, our ?nding that accounting discretion may be valuable in
performance evaluation, even absent a contracted “preferred” method, becomes
particularly relevant.
Of course, not all discretion is bene?cial in the performance evaluation sense,
particularly if there is no announced preferred accounting method. In this light, let us
recast the recent debate on one particularly contentious issue of accounting discretion:
discretion over the use of mark-to-market accounting. The debate centered on whether
mark-to-market is really informative about ?rm value when markets are frozen, and
moreover whether mark-to-market accounting can lead to a downward spiral of asset
prices (Laux and Leuz, 2009). Not to discount these issues of valuation and contagion,
but from a contracting perspective the real issue is whether adding the option to
abandon mark-to-market accounting will help evaluate the performance of managers.
This is not an easy question to answer, but note the whole notion of value relevance is
set aside. If one wants to argue that the discretion to avoid mark-to-market accounting
is desirable, then one must show that the signal obtained from the manager’s discretion
is more informative than the signal obtained from con?ning the manager to
mark-to-market.
In summary, we have illustrated that accounting discretion can be helpful in
performance evaluation and is unquestionably helpful if accounting method can be
credibly disclosed. Additional evidence from empirical studies would be useful in
determining the extent to which sophisticated boards of directors take into
consideration the potential bene?cial effects of accounting discretion.
Our paper ?ts in with the more general issue of whether uniform accounting
standards are desirable[11]. In fact, granting accounting discretion to managers
implies there are non-uniform accounting standards, but bene?ts to discretion are
much more likely if we allow the evolution of social conventions that suggest certain
accounting methods are “preferred”. This is quite different from the normal discussion
on accounting discretion.
ARJ
26,2
162
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Notes
1. In a related literature, Demski (1998), Arya et al. (1998) and Arya et al. (2003) demonstrate
that it may be ef?cient to allow managers to shift income to different periods while forcing
them to the Law of Conservation of Income (Sunder, 1997).
2. It is well established that in friction-laden markets there may not exist a well-de?ned concept
of income and value (Fellingham, 2012, chapter 5). Such a setting would render accounting
valueless if its only contribution were to estimate ?rm value!
3. There is nothing special about using signals with homogenous conditional distributions.
We could illustrate the same insights with heterogeneous signals; however doing so would
needlessly complicate the notation.
4. The word “evidence” is used somewhat metaphorically, because given the conditional wages
chosen by the owner the manager will surely choose H. That is, there is no updating of
beliefs that normally accompanies the presentation of evidence.
5. The expected likelihood ratio is always 1. For example, with no discretion it is equal to
0.7(0.714) þ 0.3[1.667] ¼ 1.
6. As noted in Penno (1990), if the number of signals were suf?ciently large, accounting
discretion necessarily becomes bene?cial. Further, as the number of signals approaches
in?nity, a ?rst-best solution is possible. The easiest way to see this is to note that as the
number of signals approaches in?nity the likelihood ratio on b approaches in?nity,
becoming perfectly informative in the limit. As a caveat, in order for this result to hold there
must be no lower bound on compensation.
7. From a purely technical standpoint it is also necessary with a square root utility function to
assume wages are non-negative.
8. Independent signals, as in illustration 6, guarantee conditional informativeness.
9. With different parameters there can be a strict gain to the owner in spite of the binding
bankruptcy constraints.
10. Demski et al. (1984) also model an agency relationship where the manager is evaluated both
on observable outcome and the accounting method chosen.
11. Sunder (2010) presents a number of advantages to allowing non-uniform accounting
standards: “The pursuit of uniform written standards at the expense of social norms
diminishes the effectiveness of ?nancial reporting in stewardship and governance, and in
keeping the security markets informed.”
References
American Accounting Association (2011), “Conceptual issues in ?nancial reporting”,
Financial Accounting Standards Committee, American Accounting Association,
Washington, DC.
Antle, R. and Demski, J. (1988), “The controllability principle in responsibility accounting”,
The Accounting Review, Vol. 63 No. 4, pp. 700-718.
Arya, A., Glover, J. and Sunder, S. (1998), “Earnings management and the revelation principle”,
Review of Accounting Studies, Vol. 3, pp. 7-34.
Arya, A., Glover, J. and Sunder, S. (2003), “Are unmanaged earnings always better for
shareholders?”, Accounting Horizons, Vol. 17, pp. 111-116.
Beatty, A. and Weber, J. (2003), “The effect of debt contracting on voluntary accounting method
changes”, The Accounting Review, Vol. 78 No. 1, pp. 119-142.
Observations
on accounting
discretion
163
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Beatty, A., Ramesh, K. and Weber, J. (2002), “The importance of accounting changes in debt
contracts: the cost of ?exibility in covenant calculations”, Journal of Accounting and
Economics, Vol. 33 No. 2, pp. 205-227.
Bebchuk, L. and Fried, J. (2003), “Executive compensation as an agency problem”, Journal of
Economic Perspectives, Vol. 17, pp. 71-92.
Benston, G. (2006), “Fair-value accounting: a cautionary tale from Enron”, Journal of Accounting
& Public Policy, Vol. 25 No. 4, pp. 465-484.
Bergstresser, D. and Phillipon, T. (2006), “CEO incentives and earnings management”, Journal of
Financial Economics, Vol. 80, pp. 511-529.
Bowen, R., Rajgopal, S. and Venkatachalam, M. (2008), “Accounting discretion, corporate
governance, and ?rm performance”, Contemporary Accounting Research, Vol. 25 No. 2,
pp. 351-405.
Briloff, A. (1972), Unaccountable Accounting, Harper & Row, New York, NY.
Chambers, D., Jennings, R. and Thompson, R. II (2003), “Managerial discretion and accounting
for research and development costs”, Journal of Accounting, Auditing & Finance, Vol. 18
No. 1, pp. 79-113.
Christensen, J. and Demski, J. (2003), Accounting Theory: An Information Content Perspective,
McGraw-Hill Irwin, New York, NY.
Christie, A. and Zimmerman, J. (1994), “Ef?cient and opportunistic choices of accounting
procedures: corporate control contests”, The Accounting Review, Vol. 69 No. 4, pp. 539-566.
Coles, J., Hertzel, M. and Kalpathy, S. (2006), “Earnings management and employee stock option
reissues”, Journal of Accounting and Economics, Vol. 41 Nos 1/2, pp. 173-200.
Dechow, P., Huson, M. and Sloan, R. (1994), “The effect of restructuring charges on executives’
cash compensation”, The Accounting Review, Vol. 69 No. 1, pp. 138-156.
Demski, J. (1998), “Performance measure manipulation”, Contemporary Accounting Research,
Vol. 15, pp. 261-285.
Demski, J., Patell, J. and Wolfson, M. (1984), “Decentralized choice of monitoring systems”,
The Accounting Review, Vol. 59 No. 1, pp. 16-34.
Fellingham, J. (2012), Accounting Structure, unpublished monograph.
Jensen, M., Murphy, K. and Wruck, E. (2004), “Remuneration: where we’ve been, how we got to
here, what are the problems, and how to ?x them”, working paper, Harvard University,
Cambridge, MA.
Kanodia, C., Singh, R. and Spero, A. (2005), “Imprecision in accounting measurement: can it be
value enhancing?”, Journal of Accounting Research, Vol. 43 No. 3, pp. 487-519.
Kim, S. and Suh, Y. (1991), “Ranking of accounting information systems for management”,
Journal of Accounting Research, Vol. 29 No. 2, pp. 386-396.
Kren, L. and Leauby, B. (2001), “The effect of SFAS 106 on chief executive compensation”,
Advances in Public Interest Accounting, Vol. 8 No. 1, pp. 143-167.
Levitt, A. (1998), “The numbers game”, Remarks delivered at the NYU Center for Law and
Business, New York, NY, September 28.
Ozbilgin, M. and Penno, M. (2008), “The assignment of decision rights in formal information
systems”, Journal of Management Accounting Research, Vol. 20, pp. 83-105.
Penno, M. (1990), “Auditing for performance evaluation”, The Accounting Review, Vol. 65 No. 3,
pp. 520-536.
Salamon, G. and Smith, E. (1979), “Corporate control and managerial misrepresentation of ?rm
performance”, The Bell Journal of Economics, Vol. 10 No. 1, pp. 319-328.
ARJ
26,2
164
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Schwarcz, S. (2002), “Enron and the use and abuse of special purpose entities in corporate
structure”, The Financier, Vol. 9, pp. 23-29.
Securities and Exchange Commission (2001), Securities Exchange Act of 1934, Release No. 44444,
Securities and Exchange Commission, available at: www.sec.gov/litigation/admin/
34-44444.htm
Sunder, S. (1997), Theory of Accounting and Control, South-Western, Cincinnati, OH.
Sunder, S. (2010), “Adverse effects of uniform written reporting standards on accounting practice,
education, and research”, Journal of Accounting & Public Policy, Vol. 29, pp. 99-114.
Tof?er, B. and Reingold, J. (2003), Final Accounting: Ambition, Greed, and the Fall of Arthur
Andersen, Broadway Books, New York, NY.
Watts, R. and Zimmerman, J. (1990), “Positive accounting theory: a ten year perspective”,
The Accounting Review, Vol. 65, pp. 31-156.
Further reading
Klein, A. (2002), “Audit committee, board of director characteristics, and earnings management”,
Journal of Accounting and Economics, Vol. 33, pp. 375-400.
Appendix 1. Illustrations 1-6
Below is the mathematical program that applies for Illustrations 1 through 6 where the report is
limited to g or b. No discretion is indicated by n ¼ 1, and discretion by n $ 2 (Table AI):
w
g
;w
b
min p
g
w
g
þ ð1 2p
g
Þw
b
subject to:
p
g
??????
w
g
p
þ ð1 2p
g
Þ
?????
w
b
p
2 c
H
$
U ðPÞ
p
g
??????
w
g
p
þ ð1 2p
g
Þ
?????
w
b
p
2 c
H
$ q
g
??????
w
g
p
þ ð1 2q
g
Þ
?????
w
b
p
ðICÞ
w
g
; w
b
; $ 0 ðBÞ
where n is the number of available signals, and:
p
g
¼ 1 2ð1 2pÞ
n
and q
g
¼ 1 2ð1 2qÞ
n
p ¼ probability of a signal of g, given H.
q ¼ probability of a signal of g, given L.
p
g
¼ probability manager reports g, given H.
q
g
¼ probability manager reports g, given L.
Illustration n p q p
g
q
g
c
H
U
¯
1 1 0.7 0.5 0.7 0.5 50 500
2 2 0.7 0.5 0.91 0.75 50 500
3 1 0.97 0.96 0.97 0.96 1 2,500
4 2 0.97 0.96 0.9991 0.9984 1 2,500
5 1 0.7 0.5 0.7 0.5 50 50
6 2 0.7 0.5 0.91 0.75 50 50
Table AI.
Observations
on accounting
discretion
165
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Appendix 2. Illustrations 7-8
Below are the programs that apply to Illustrations 7 and 8, in which the manager has been
granted discretion to choose from between two signals and the owner declares a preferred signal:
w
pg
;w
ng
;w
b
min p
pg
w
pg
þ p
ng
w
ng
þ p
b
w
b
subject to:
p
pg
???????
w
pg
p
þ p
ng
???????
w
ng
p
þ p
b
?????
w
b
p
2 c
H
$
U ðPÞ
p
pg
???????
w
pg
p
þ p
ng
???????
w
ng
p
þ p
b
?????
w
b
p
2 c
H
$ q
pg
???????
w
pg
p
þ q
ng
???????
w
ng
p
þ q
b
?????
w
b
p
ðICÞ
w
pg
; w
ng
w
b
; $ 0 ðBÞ
w
pg
¼ payment for a report of g from the preferred signal.
w
ng
¼ payment for a report of g from the non-preferred signal.
w
b
¼ payment for a report of b.
Probabilities given action H are: p
pg
¼ p, p
np
¼ 1 2ð1 2pÞ
2
2p, p
b
¼ 1 2½1 2ð1 2pÞ
2
?.
Probabilities given action L are: q
pg
¼ q q
np
¼ 1 2 (1 2 q)
2
2 q, q
b
¼ 1 2 [1 2 (1 2 q)
2
].
Illustration 7 (based on 3 and 4): p ¼ 0.97, q ¼ 0.96,U
¯
¼ 2,500, and c
H
¼ 1.
Illustration 8 (based on 5 and 6): p ¼ 0.7, q ¼ 0.5, U
¯
¼ 50, and c
H
¼ 50.
Corresponding author
Steven Schwartz can be contacted at: [email protected]
ARJ
26,2
166
To purchase reprints of this article please e-mail: [email protected]
Or visit our web site for further details: www.emeraldinsight.com/reprints
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_425370728.pdf
The purpose of this paper is to explore the role of accounting discretion in a
principal-agent setting, wherein accounting information is used in performance evaluation. The agent
may choose one from among several allowable accounting methods. However, limited audit resources
allow only for verification of only the method the agent chooses, and this is the only one used to
determine the agent’s compensation.
Accounting Research Journal
A few stylized observations on accounting discretion
Steven Schwartz Richard Young
Article information:
To cite this document:
Steven Schwartz Richard Young , (2013),"A few stylized observations on accounting discretion",
Accounting Research J ournal, Vol. 26 Iss 2 pp. 154 - 166
Permanent link to this document:http://dx.doi.org/10.1108/ARJ -J ul-2012-0064
Downloaded on: 24 January 2016, At: 21:17 (PT)
References: this document contains references to 34 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 482 times since 2013*
Users who downloaded this article also downloaded:
Amanda J . Carter, Roger L. Burritt, J ohn D. Pisaniello, (2013),"The dual roles for accountants in sustaining
rural communities", Accounting Research J ournal, Vol. 26 Iss 2 pp. 130-153http://dx.doi.org/10.1108/
ARJ -12-2012-0093
Victoria J . Clout, Larelle Chapple, Nilan Gandhi, (2013),"The impact of auditor independence regulations
on established and emerging firms", Accounting Research J ournal, Vol. 26 Iss 2 pp. 88-108 http://
dx.doi.org/10.1108/ARJ -Dec-2011-0045
Vijaya Murthy, J ames Guthrie, (2013),"Accounting for workplace flexibility: Internal communication
in an Australian financial institution", Accounting Research J ournal, Vol. 26 Iss 2 pp. 109-129 http://
dx.doi.org/10.1108/ARJ -12-2012-0096
Access to this document was granted through an Emerald subscription provided by emerald-srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.
*Related content and download information correct at time of download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
A few stylized observations
on accounting discretion
Steven Schwartz
School of Management, Binghamton University, Binghamton,
New York, USA, and
Richard Young
Fisher College of Business, The Ohio State University,
Columbus, Ohio, USA
Abstract
Purpose – The purpose of this paper is to explore the role of accounting discretion in a
principal-agent setting, wherein accounting information is used in performance evaluation. The agent
may choose one from among several allowable accounting methods. However, limited audit resources
allow only for veri?cation of only the method the agent chooses, and this is the only one used to
determine the agent’s compensation.
Design/methodology/approach – This paper uses the model developed by Penno (1990) and
construct relatively simple examples that illustrate the effects of accounting discretion on performance
evaluation. These examples may make the implications of Penno (1990) clearer to educators and
practitioners.
Findings – Accounting discretion can be useful in performance evaluation even if the accounting
method chosen is not disclosed. Lower limits on compensation reduce the usefulness of accounting
discretion. If the principal declares a preferred accounting treatment, it enhances the usefulness of
accounting discretion; in fact, it guarantees that accounting discretion is weakly preferred to no discretion.
Practical implications – Accounting discretion is often viewed pejoratively. Those responsible for
designing performance evaluation schemes should be aware of the potential bene?ts of accounting
discretion, and those charged with promulgating accounting rules should consider the possibility that
rules can be too in?exible.
Originality/value – This paper provides accessible illustrations of the effects of accounting
discretion on performance evaluation and offer discussion to place the conclusions drawn from the
model into a standard setting context.
Keywords Performance evaluation, Agency theory, Accounting discretion, Performance appraisal,
Accountancy
Paper type Research paper
Managers clearly have discretion in the way they report their ?rms’ ?nancial results.
Such discretion is often portrayed in the popular press as undesirable (Briloff, 1972;
Levitt, 1998). In many cases, this view is justi?ed. For example, managers at Enron took
liberal advantage of the rules regarding mark-to-market accounting and the
consolidation of special purpose entities (Schwarcz, 2002), while managers at waste
management exploited accounting rules regarding changes in the useful life of
depreciable assets. With the bene?t of hindsight, it is clear both ?rms went beyond the
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1030-9616.htm
The authors would like to thank participants at the Northeast Regional Meeting of the American
Accounting Association and The Ohio State University Accounting Honors students for their
helpful comments and suggestions.
Accounting Research Journal
Vol. 26 No. 2, 2013
pp. 154-166
qEmerald Group Publishing Limited
1030-9616
DOI 10.1108/ARJ-Jul-2012-0064
ARJ
26,2
154
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
limits allowed by generally accepted accounting principles (GAAP). Yet there was
enough discretion allowed in GAAP for each ?rms’ ?nancial reports to receive a clean
bill of health from their auditor, Arthur Andersen, despite the auditor’s general
knowledge of what the ?rm had done (Benston, 2006; SEC, 2001; Tof?er and Reingold,
2003). The academic literature also tends to view accounting discretion pejoratively,
mainly as a means by which managers can expropriate wealth from the ?rm. Examples
include managers using accounting discretion so as to maximize the value of their
holdings of equity options or to ward off takeover attempts that might otherwise bene?t
shareholders (Bergstresser andPhillipon, 2006; Christie and Zimmerman, 1994; Salamon
and Smith, 1979).
We describe and illustrate an alternative view of discretion, using the stylized model
of Penno (1990). We demonstrate that when a principal, such as a board of directors,
grants accounting discretion to an agent, such as a CEO, it may decrease agency costs.
Underlying our analysis is an assumption that sophisticated boards of directors, when
devising incentive contracts, anticipate how managers will use their discretion
(Bowen et al., 2008; Watts and Zimmerman, 1990). Given that boards do so, optimal
contracts that allow for accounting discretion can be bene?cial. Consideration of this
possibility is important not only from the perspective of contract design, but also
policy decisions on the ?exibility of accounting standards.
Penno (1990) models accounting discretion as the ability of the manager to choose
one signal from a vector of signals with common knowledge probability distribution.
The signal chosen cannot be misrepresented (the so-called “anti-fraud rule”), but only
the manager knows the value of the other signals. Penno shows that allowing an agent
a choice of which signal to report provides contracting bene?ts to the principal, despite
the fact that the agent will choose the signal that puts him in the best light. Ozbilgin
and Penno (2008) explore the conditions under which the owner wishes to delegate the
choice of signal to the manager or retain the decision rights to choose the signal[1].
We use the Penno (1990) framework to illustrate four stylized observations about
accounting discretion in a performance evaluation setting. First, accounting discretion
may be bene?cial even if a manager discloses neither his choice of the accounting
methodnor the outcome of the accounting methods he did not use. Second, a lower bound
on a manager’s compensation decreases the effectiveness of accounting discretion in
performance evaluation. Third, if a manager is required to disclose his chosen
accounting method and there is no lower limit on compensation, accounting discretion is
strictly preferred. Fourth, if a manager is required to disclose his chosen accounting
method, a lower bound on compensation may diminish the bene?t of accounting
discretion, but nevertheless more discretion is still weakly preferred to less.
Unlike the literature on accounting disclosure and equity pricing, from a contracting
perspective accounting discretion can be seen as useful under a robust set of
circumstances. In addition, the arguments we make herein do not rely on the notion
that accountants need discretion in order to better communicate the value of the ?rm
(Chambers et al., 2003). In fact, despite what seems to be implied by the language of
accounting, interpreting accounting numbers as economic value and income can be
troublesome in a world where markets are not frictionless. “Therefore, it is important
to look at the usefulness of accounting information as being something other than an
input in the estimation of ?rm value” (See, for example, Christensen and Demski, 2003;
Kanodia et al., 2005)[2]. We end with a brief look at one form of accounting based
Observations
on accounting
discretion
155
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
contracts, debt covenants. We note that creditors often allow debtors accounting
discretion although contracts could be written otherwise.
Model
No reporting discretion (one signal)
Consider a simple binary action agency model, as in Antle and Demski (1988). A ?rm
consists of an owner (principal) and a manager (agent) who contract over a single task.
The manager chooses an action a, where a [{H, L}. The manager has a private cost of
H but no private cost for choosing L: c
H
. c
L
¼ 0. We assume action H is suf?ciently
bene?cial to the owner that he wishes to motivate H, even though he will have to pay
the manager a greater amount for H than L.
The action chosen by the manager is not observable to the owner and so cannot be
contracted upon directly. However, the owner and manager can contract upon a signal
that the manager can in?uence. The signal may be interpreted as the product of the
accounting system. The manager’s report of the signal is veri?ed without error by an
auditor. Assuming a suf?cient penalty is available for a report that is inconsistent with
the auditor’s ?ndings, the manager will in equilibrium report the signal truthfully. For
simplicity, assume the signal has a binary outcome that is either g (good) or b (bad). Let
p be the probability of g given a ¼ H and q be the probability of g given a ¼ L, where
p . q, justifying the labels of “good” and “bad” for g and b, respectively.
The manager is risk-averse and has a utility for money equal to
????
w
p
. In the
contracting program the risk-neutral owner minimizes the expected wage paid to
the manager subject to two constraints. The “participation” constraint (P) guarantees
the manager an expected utility of at least
U, which is the expected utility obtained if
he went elsewhere. The “incentive compatibility” constraint (IC) ensures the choice of H
is weakly better for the manager than L.
The risk neutral owner’s problem to ef?ciently motivate H is found below:
min
w
g
;w
b
pw
g
þ ð1 2pÞw
b
subject to:
p
??????
w
g
p
þ ð1 2pÞ
?????
w
b
p
2 c
H
$
U ðPÞ
p
??????
w
g
p
þ ð1 2pÞ
?????
w
b
p
2 c
H
$ q
??????
w
g
p
þ ð1 2qÞ
?????
w
b
p
ðICÞ
In illustration 1, we set p ¼ 0.7, q ¼ 0.5, c
H
¼ 50 and
U ¼ 500. The solution, found in
Table I, is w
g
¼ 390, 625, w
b
¼ 140,625 and E[w] ¼ 315,625. We can think of the
contract as a base wage of (c
H
þ U)
2
¼ 302,500, which compensates the manager
for the disutility of H and for choosing to participate in the ?rm, with a bonus of 88,125
for a good report and a penalty of 161,875 for a bad report. The bonus and penalty are
necessary to motivate the desired action by the manager when the action is not directly
observable. The expected wage of 315,625 is greater than base wage of 302,500 which
re?ects the need to compensate the risk-averse manager for receiving an uncertain
wage. An improved information system would reduce the amount of risk placed on the
manager; this is what we search for when we introduce accounting discretion.
ARJ
26,2
156
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Accounting discretion
In the previous sub-section the accounting system produced only one signal, to model
the idea of no discretion. Now assume the accounting system generates two
independently distributed signals. The different signals canbe thought of as the result of
different allowable accounting methods, for example FIFO and LIFO. The manager, at
his own discretion, submits a report that must be equal to one of the two signals he
privately observes. The chosen signal cannot be misreported, but perhaps due to limited
auditing resources, only the chosen signal will be veri?ed and reported to the owner.
In this section the agent cannot credibly disclose which of the two signals he chose,
so upon a report of g, the owner only knows that at least one of the two signals is g.
With these assumptions g continues to be good news about the manager’s action
choice. Therefore, the manager will choose to report g, whenever possible, implying a
report of b indicates both signals are b. We maintain the assumptions of p ¼ 0.7 and
q ¼ 0.5, so that the probability of being capable of reporting g is equal to
1 2 (1 2 p)
2
¼ 0.91 conditional on H and 1 2 (1 2 q)
2
¼ 0.75 conditional on L[3]. The
owner’s program for this and subsequent illustrations is found in the Appendix, and
the solution to the program is shown in Table I as illustration 2.
From Table I we observe the cost of motivating H has decreased from 315,625 to
310,498 with the introduction of accounting discretion. This is one of the primary
insights in Penno (1990) – accounting discretion can be valuable from a contracting
perspective. The key to understanding this result is to examine the likelihood ratios that
correspond to the two feasible reports (Antle and Demski, 1988). The likelihood ratio is
the probability of a particular report conditional on L divided by the probability of the
same report conditional on H. A likelihood ratio greater than 1 is “evidence” that L was
taken; a likelihood ratio less than 1 is evidence H was taken[4]. A likelihood ratio of
exactly 1 would imply that observing that particular report is not helpful in determining
which action was taken.
When discretion is added in this example the likelihood ratio for a report of
g moves closer to 1 (from 0.714 to 0.824) and the bonus decreases to
(334,228.5 2 302,500) ¼ 31,728.5. As the likelihood ratio moves closer to 1 the evidence
that H was taken becomes weaker, so it is ef?cient for the owner to reduce the bonus.
Intuitively, increasing the number of potential signals increases the probability g will
be reported regardless of the action taken, hence the bonus decreases.
Illustration 1 Illustration 2 Illustration 3 Illustration 4
Number of signals 1 2 1 2
Conditional
distribution p ¼ 0.7; q ¼ 0.5 p ¼ 0.7; q ¼ 0.5 p ¼ 0.97; q ¼ 0.96 p ¼ 0.97; q ¼ 0.96
Parameters c
H
¼ 50;
U ¼ 500 c
H
¼ 50;
U ¼ 500 c
H
¼ 1;
U ¼ 2,500 c
H
¼ 1;
U ¼ 2,500
Pr(report ¼ gjL) 0.5 0.75 0.96 0.9984
Pr(report ¼ gjH) 0.7 0.91 0.97 0.9991
LR(g) 0.714 0.824 0.9897 0.9993
LR(b) 1.667 2.78 1.3333 1.7778
w
g
390,625 334,228.5 6,270,016 6,261,433
w
b
140,625 70,556.4 5,779,216 1,152,862
E[Wage] 315,625 310,498 6,255,292 6,256,836
Var[LR] 0.190 0.313 0.0034 0.0005
Table I.
Results for illustrations
1, 2, 3 and 4
Observations
on accounting
discretion
157
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Conversely, inmovingfromone signal to two signals the penalty for a report of b increases
to (302,500 2 70,556.4) ¼ 231,943.6, because the likelihood ratio moves further from
1 (from 1.67 to 2.78). Intuitively, increasing the number of potential signals decreases the
likelihood that b will be reported regardless of the action taken, hence the penalty for a
report of b increases.
Ideally, we would like both likelihood ratios to be as far from their expected value
(equal to 1) as possible. Therefore, it is perhaps not surprising that the greater the
“spread” of the likelihood ratios, the more valuable the information systemin evaluating
the manager. Two likelihood ratios of 1 would indicate a system with no performance
evaluation information. Two likelihood ratios that approach 0 and in?nity, for the good
and bad signal, respectively, would indicate a system that provides near-perfect
information. More formally, with a square root utility function for the manager the value
of the accounting system is strictly increasing in the variance of the likelihood ratios
(Kim and Suh, 1991). For the case of no discretion, the variance of the likelihood ratios is
0.7(0.714 2 1)
2
þ 0.3(1.667 2 1)
2
¼ 0.190[5]. For the case of discretion, the likelihood
ratio is 0.313 . 0.190. Comparing the expected wages in illustrations 1 and 2, the owner
prefers to allow discretion on the part of the manager.
So far we have shown that accounting discretion may be valuable even if the owner is
unaware of the accounting method chosen. However, this result is not general. Let us
alter the conditional probability of the signals so that p ¼ 0.97 and q ¼ 0.96. Further let
us assume that c
H
¼ 1 and
U ¼ 2,500. With these new parameters, the no discretion
case is shown in Table I as illustration 3 and the two signals case is shown in Table I as
illustration 4. Our reason for presenting illustrations 3 and 4 is immediate: whether
accounting discretion is bene?cial depends on the details. With the parameters in
illustrations 3 and 4, expected wages increase when discretion is added. As mentioned,
there are countervailing forces at work when accounting discretion is introduced.
A signal of b becomes more informative while a signal g becomes less informative. In
illustration 4 the loss of information fromg outweighs the gain of information fromb[6].
The setting we have examined so far is not uncommon in ?nancial reporting.
Although ?nancial statements contain considerable documentation on accounting
method choice, clearly not all choices are transparent. Yet, we rely on auditors for
assurances that all choices ultimately fall within the realm of GAAP. That is, we know
(or hope) that the report comes from an allowable method but we may not be aware of
exactly the method chosen. The fact that accounting discretion can still be valuable
under these assumptions is both surprising and important in explaining the wide
discretion available in ?nancial reporting.
Bankruptcy constraints
In almost all realistic settings involving performance evaluation and compensation
there is a lower bound on employee remuneration. One useful representation of this
lower bound is a bankruptcy constraint such that employees must receive a
non-negative wage. This makes sense because employees may be ?red but typically do
not pay their ?rm a penalty that exceeds their wages[7]. We explore the effect of a
bankruptcy constraint with illustrations 5 and 6.
In illustration 5, we assume a single signal with the conditional distributions
used in illustrations 1 and 2: p ¼ 0.7 and q ¼ 0.5. However, we change
U to
50 while maintaining c
H
¼ 50. This has the effect of making a bankruptcy constraint
ARJ
26,2
158
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
on w
b
binding. The results are shown in Table II, where we observe the manager
receives 62,500 for g and 0 for b. In comparison, the ?rst-best solution (no information
asymmetry) would pay (c
H
þ U)
2
¼ 10,000, so we may interpret the contract as paying
the manager a bonus of 52,500 for a good outcome and a penalty of 10,000 for a bad
outcome. Recall, when
U ¼ 500 the penalty was almost twice the size of the bonus
(161,875 compared to 88,125). As one might surmise, the owner would like to implement
a similar ratio of penalty to bonus in illustration 5, but cannot due to the bankruptcy
constraint. It should come as no surprise then that as we increase the informativeness of
b and decrease the informativeness of g through accounting discretion, satisfying the
bankruptcy constraint will become increasingly costly and the owner will be made
worse off with discretion. As an analogy, imagine the criminal justice system worked
only be giving those citizens who have been “proven” to be innocent a bonus, with the
further constraint that most people would appear innocent regardless of whether they
had committed a crime. Such a system would be both costly and inef?cient.
In illustration 6 the manager has discretion over two signals with the same
parameters as in illustration 5. The payment for a good report is now97,656.25 while the
payment for a bad report remains at 0. Relative to the no discretion case, expected wages
increase from43,750 to 88,867. The wage for a bad report was already at the lower bound
of zero in illustration 5. Adding a second signal in illustration 6 causes the good outcome
to become even less informative, so the ef?cient contract must increase the bonus even
further (with no ability to increase the penalty) and expected compensation increases.
The overarching story here is that in a performance evaluation setting a lower bound on
compensation can make accounting discretion particularly unfavorable for the owner.
Two observations are in order. As noted, in almost all practical situations
compensation is bounded from below. Most employees cannot be forced to pay back
money to the ?rm except in cases of theft, so that dismissal would be the strongest
penalty. For CEOs and other top executives, even dismissal may be very pecunious
(Bebchuk and Fried, 2003). Therefore, the usefulness of discretion may seem to be
limited, due to restrictions on penalties that boards of directors can realistically impose.
Second, Ozbilgin and Penno (2008) demonstrate that when there is a lower bound on
compensation discretion may still be useful, but only in the case where it is optimal to
give discretion to the owner. With the owner retaining discretion over which signal is
chosen, the bankruptcy constraints for the manager would not be problematic. The
owner would choose the signal that corresponds to the lowest wage, if possible. Hence,
report of a g implies all signals are g and further implies g becomes the more
informative signal. In consequence a report of g is accompanied with a large bonus
while a report of b is accompanied by a small penalty, which is helpful when the
bankruptcy constraint is binding.
Illustration 5: one signal Illustration 6: two signals
Pr(gjL) 0.5 0.75
Pr(gjH) 0.7 0.91
w
g
62,500 97,656.25
w
b
0 0
E[Wage] 43,750 88,867.19
Table II.
Likelihood ratios (LR) for
illustrations 5 and 6
p ¼ 0.7; q ¼ 0.5;
c
H
¼ 50;
U ¼ 50
Observations
on accounting
discretion
159
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Accounting preferences and customs
To this point we have assumed the signals are indistinguishable. We now assume the
manager can credibly disclose which of the signals is being reported. This is in essence
what occurs when the manager provides details in the footnotes about the particular
accounting method chosen. We maintain the assumption that the manager only reports
one signal, and does so in his own best interest.
Let us ?rst return to the parameters used in illustrations 3 and 4. We demonstrated
earlier the owner would be better off if she only allowed the manager only a single
signal. Suppose now that the owner establishes a preference for the manager to report
from one of the two signals. While it is not valuable to force the manager to divulge
which of the two signals he is reporting per se, it is interesting that it is valuable if the
owner establishes a preference ex ante. The reason is simple: if the non-preferred signal
is disclosed as g it can be inferred that the preferred signal is b. Clearly, the owner
potentially receives an additional piece of information if he announces a preference for
the use of a particular signal, relative to the case of no preference. We demonstrate with
illustration 7, found in Table III.
In illustration 7, we label one signal the preferred signal; the other the non-preferred
signal. The probability the manager reports g from the preferred signal conditional on
action H is 0.97, while the probability that the manager reports g from the non-preferred
system conditional on action H is (1 2 0.97)0.03 ¼ 0.0291. Finally, the probability the
manager cannot report g conditional on His equal to 0.03(0.03) ¼ 0.0009. The solution to
illustration 7 is provided in Table III, where w
pg
and w
ng
are the wages for a report of g
from the preferred and non-preferred signals, respectively. The program to solve for the
optimal contract for this illustration (and illustration 8) is in Appendix 2. The expected
wage payment has decreased from 6,255,292 with a single signal (no discretion) to
6,255,277 with two signals and granting discretion to the manager.
Illustration 7 re?ects the general result that as long as the owner announces a
preference for a particular preference, and with no binding bankruptcy constraint, the
owner is strictly better off increasing discretion by adding a new system, as long as the
new system is informative given the ?rst[8]. The intuition is as follows. If the owner
sets a preference, she will always know if the signal coming from the preferred signal is
g or b, as in the case of a single signal. However, she will also know whether the
non-preferred signal is g or b, given the preferred signal is b. Technically speaking,
there is an additional partition of the information space. Hence, adding an additional
Illustration 7: p ¼ 0.97, q ¼ 0.96,
c
H
¼ 1,
U ¼ 2,500
Illustration 8: p ¼ 0.7, q ¼ 0.5,
c
H
¼ 50,
U ¼ 50
Pr( pgjL) 0.96 0.5
Pr( pgjH) 0.97 0.7
Pr(ngjL) 0.0384 0.25
Pr(ngjH) 0.0291 0.21
w
pg
6,269,248 62,500
w
ng
5,821,566 0
w
b
5,221,210 0
E[Wage] 6,255,277 43,750
Table III.
Illustrations 7 and 8 –
accounting discretion
with a preferred signal
ARJ
26,2
160
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
system cannot be harmful to the owner, and without a binding bankruptcy constraint it
is strictly bene?cial. Note the very robust result here.
In illustration 8 we continue to assume that the owner has a preference, but we return
to the parameters from illustrations 5 and 6. Recall, in illustrations 5 and 6 accounting
discretion was harmful to the owner due to the lower bound on compensation to the
manager. Under an optimal contract, due to the binding bankruptcy constraint, w
pg
is
the only non-zero wage. The expected wage of 43,750 is equal to that resulting from a
single signal. In essence the contract replicates the no-discretion contract, the owner does
not gain, but does not lose either. As with illustration 7, the preference causes a partition
in the information space, but in this instance there is no bene?t to using it[9]. This
example illustrates the general result that as long as a preference is established up front
by the owner, there is no information lost adding a new signal, and hence even with a
bankruptcy constraint, accounting discretion is weakly preferred.
It is interesting that boards of directors do not typically state preferred accounting
methods in their executive compensation schemes. Also, market based compensation,
such as employee stock options, has become the dominant form of executive
compensation ( Jensen et al., 2004). This perhaps is where the value of accounting
conventions or customs comes into play. A custom can be thought of as a habitual
practice or behavior that people engage in and expect others to engage in. Customs can
have social bene?t. For example, farmers’ markets being held after certain holidays
have a coordination bene?t. Similarly, when discussing accounting conventions the
American Accounting Association’s Financial Accounting Standards Committee noted
“[accounting conventions] may be valued indirectly for their coordination bene?ts that
even arbitrary choices yield through their social acceptance and wide use”
(American Accounting Association, 2011, p. 11). The market might understand
which accounting method is conventional, and so would react negatively if a method
that is technically acceptable, but not conventional, were used, This phenomenon is
similar to the idea of a preference in our model, in that the market is rightly inferring a
worse report would have resulted had the conventional method been used. Ultimately,
depending on the circumstances, allowing the manager discretion may be ex ante
ef?cient from a contracting perspective.
Discussion and conclusion
Accounting is fundamentally a source of information (Christensen and Demski, 2003).
Accounting information may be useful in valuing a ?rm, determining a ?rms’ tax
liability or evaluating whether a ?rm has met regulatory requirements. Therefore, in
selecting an accounting information regime one should be mindful of its purpose.
In this paper our focus is on accounting’s role in providing useful information for
performance evaluation. We illustrate that accounting discretion can be useful in this
endeavor. One natural question that arises though with respect to our analyses is: do
contract designers, such as boards of directors, consider the effects of accounting
discretion when designing contracts? This may seem like lot to ask from Board
members. However, Coles et al. (2006) provide evidence that market participants
anticipate earnings management around the re-issuance of stock options, so it may not
be so unreasonable that boards of directors anticipate the behavior as well. The situation
we model is similar to accounting discretion and accounting-based debt covenants. We
know a lot about debt covenants, so it might be fruitful to look there for contracting
Observations
on accounting
discretion
161
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
evidence related to accounting discretion. Creditors can easily write contracts restricting
discretion over accounting method and indeed charge lower interest rates when doing so
(Beatty et al., 2002). However, creditors generally allow discretion, and research has
shown that debtors use the discretion to avoid default (Beatty and Weber, 2003). One
possibility is that for some creditor-debtor relationships accounting discretion is part of
an ef?cient contract.
Of more direct relevance is whether accounting based performance evaluation
schemes consider the accounting choices of managers[10]. There is some evidence that
managers are evaluated differently based on how they report income. For example,
non-recurring charges and losses due to forced changes in accounting procedure tend to
not affect the manager’s compensation (Dechow et al., 1994;Kren and Leauby, 2001).
However, these are really measures to protect the manager against circumstances
beyond his control, rather thanto encourage the use of a particular accounting treatment.
It has been pointed out that it may be costly to proscribe certain accounting treatments,
because managers have a better understanding of which accounting methods are most
bene?cial for the ?rm from an overall perspective (e.g. in aiding make-or-buy decisions
and pricing decisions) rather than from just a performance evaluation perspective
(Christie and Zimmerman, 1994). In fact, Bowen et al. (2008) provide evidence that
greater accounting discretion may be associated with better ?rm performance in the
future. Therefore, our ?nding that accounting discretion may be valuable in
performance evaluation, even absent a contracted “preferred” method, becomes
particularly relevant.
Of course, not all discretion is bene?cial in the performance evaluation sense,
particularly if there is no announced preferred accounting method. In this light, let us
recast the recent debate on one particularly contentious issue of accounting discretion:
discretion over the use of mark-to-market accounting. The debate centered on whether
mark-to-market is really informative about ?rm value when markets are frozen, and
moreover whether mark-to-market accounting can lead to a downward spiral of asset
prices (Laux and Leuz, 2009). Not to discount these issues of valuation and contagion,
but from a contracting perspective the real issue is whether adding the option to
abandon mark-to-market accounting will help evaluate the performance of managers.
This is not an easy question to answer, but note the whole notion of value relevance is
set aside. If one wants to argue that the discretion to avoid mark-to-market accounting
is desirable, then one must show that the signal obtained from the manager’s discretion
is more informative than the signal obtained from con?ning the manager to
mark-to-market.
In summary, we have illustrated that accounting discretion can be helpful in
performance evaluation and is unquestionably helpful if accounting method can be
credibly disclosed. Additional evidence from empirical studies would be useful in
determining the extent to which sophisticated boards of directors take into
consideration the potential bene?cial effects of accounting discretion.
Our paper ?ts in with the more general issue of whether uniform accounting
standards are desirable[11]. In fact, granting accounting discretion to managers
implies there are non-uniform accounting standards, but bene?ts to discretion are
much more likely if we allow the evolution of social conventions that suggest certain
accounting methods are “preferred”. This is quite different from the normal discussion
on accounting discretion.
ARJ
26,2
162
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Notes
1. In a related literature, Demski (1998), Arya et al. (1998) and Arya et al. (2003) demonstrate
that it may be ef?cient to allow managers to shift income to different periods while forcing
them to the Law of Conservation of Income (Sunder, 1997).
2. It is well established that in friction-laden markets there may not exist a well-de?ned concept
of income and value (Fellingham, 2012, chapter 5). Such a setting would render accounting
valueless if its only contribution were to estimate ?rm value!
3. There is nothing special about using signals with homogenous conditional distributions.
We could illustrate the same insights with heterogeneous signals; however doing so would
needlessly complicate the notation.
4. The word “evidence” is used somewhat metaphorically, because given the conditional wages
chosen by the owner the manager will surely choose H. That is, there is no updating of
beliefs that normally accompanies the presentation of evidence.
5. The expected likelihood ratio is always 1. For example, with no discretion it is equal to
0.7(0.714) þ 0.3[1.667] ¼ 1.
6. As noted in Penno (1990), if the number of signals were suf?ciently large, accounting
discretion necessarily becomes bene?cial. Further, as the number of signals approaches
in?nity, a ?rst-best solution is possible. The easiest way to see this is to note that as the
number of signals approaches in?nity the likelihood ratio on b approaches in?nity,
becoming perfectly informative in the limit. As a caveat, in order for this result to hold there
must be no lower bound on compensation.
7. From a purely technical standpoint it is also necessary with a square root utility function to
assume wages are non-negative.
8. Independent signals, as in illustration 6, guarantee conditional informativeness.
9. With different parameters there can be a strict gain to the owner in spite of the binding
bankruptcy constraints.
10. Demski et al. (1984) also model an agency relationship where the manager is evaluated both
on observable outcome and the accounting method chosen.
11. Sunder (2010) presents a number of advantages to allowing non-uniform accounting
standards: “The pursuit of uniform written standards at the expense of social norms
diminishes the effectiveness of ?nancial reporting in stewardship and governance, and in
keeping the security markets informed.”
References
American Accounting Association (2011), “Conceptual issues in ?nancial reporting”,
Financial Accounting Standards Committee, American Accounting Association,
Washington, DC.
Antle, R. and Demski, J. (1988), “The controllability principle in responsibility accounting”,
The Accounting Review, Vol. 63 No. 4, pp. 700-718.
Arya, A., Glover, J. and Sunder, S. (1998), “Earnings management and the revelation principle”,
Review of Accounting Studies, Vol. 3, pp. 7-34.
Arya, A., Glover, J. and Sunder, S. (2003), “Are unmanaged earnings always better for
shareholders?”, Accounting Horizons, Vol. 17, pp. 111-116.
Beatty, A. and Weber, J. (2003), “The effect of debt contracting on voluntary accounting method
changes”, The Accounting Review, Vol. 78 No. 1, pp. 119-142.
Observations
on accounting
discretion
163
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Beatty, A., Ramesh, K. and Weber, J. (2002), “The importance of accounting changes in debt
contracts: the cost of ?exibility in covenant calculations”, Journal of Accounting and
Economics, Vol. 33 No. 2, pp. 205-227.
Bebchuk, L. and Fried, J. (2003), “Executive compensation as an agency problem”, Journal of
Economic Perspectives, Vol. 17, pp. 71-92.
Benston, G. (2006), “Fair-value accounting: a cautionary tale from Enron”, Journal of Accounting
& Public Policy, Vol. 25 No. 4, pp. 465-484.
Bergstresser, D. and Phillipon, T. (2006), “CEO incentives and earnings management”, Journal of
Financial Economics, Vol. 80, pp. 511-529.
Bowen, R., Rajgopal, S. and Venkatachalam, M. (2008), “Accounting discretion, corporate
governance, and ?rm performance”, Contemporary Accounting Research, Vol. 25 No. 2,
pp. 351-405.
Briloff, A. (1972), Unaccountable Accounting, Harper & Row, New York, NY.
Chambers, D., Jennings, R. and Thompson, R. II (2003), “Managerial discretion and accounting
for research and development costs”, Journal of Accounting, Auditing & Finance, Vol. 18
No. 1, pp. 79-113.
Christensen, J. and Demski, J. (2003), Accounting Theory: An Information Content Perspective,
McGraw-Hill Irwin, New York, NY.
Christie, A. and Zimmerman, J. (1994), “Ef?cient and opportunistic choices of accounting
procedures: corporate control contests”, The Accounting Review, Vol. 69 No. 4, pp. 539-566.
Coles, J., Hertzel, M. and Kalpathy, S. (2006), “Earnings management and employee stock option
reissues”, Journal of Accounting and Economics, Vol. 41 Nos 1/2, pp. 173-200.
Dechow, P., Huson, M. and Sloan, R. (1994), “The effect of restructuring charges on executives’
cash compensation”, The Accounting Review, Vol. 69 No. 1, pp. 138-156.
Demski, J. (1998), “Performance measure manipulation”, Contemporary Accounting Research,
Vol. 15, pp. 261-285.
Demski, J., Patell, J. and Wolfson, M. (1984), “Decentralized choice of monitoring systems”,
The Accounting Review, Vol. 59 No. 1, pp. 16-34.
Fellingham, J. (2012), Accounting Structure, unpublished monograph.
Jensen, M., Murphy, K. and Wruck, E. (2004), “Remuneration: where we’ve been, how we got to
here, what are the problems, and how to ?x them”, working paper, Harvard University,
Cambridge, MA.
Kanodia, C., Singh, R. and Spero, A. (2005), “Imprecision in accounting measurement: can it be
value enhancing?”, Journal of Accounting Research, Vol. 43 No. 3, pp. 487-519.
Kim, S. and Suh, Y. (1991), “Ranking of accounting information systems for management”,
Journal of Accounting Research, Vol. 29 No. 2, pp. 386-396.
Kren, L. and Leauby, B. (2001), “The effect of SFAS 106 on chief executive compensation”,
Advances in Public Interest Accounting, Vol. 8 No. 1, pp. 143-167.
Levitt, A. (1998), “The numbers game”, Remarks delivered at the NYU Center for Law and
Business, New York, NY, September 28.
Ozbilgin, M. and Penno, M. (2008), “The assignment of decision rights in formal information
systems”, Journal of Management Accounting Research, Vol. 20, pp. 83-105.
Penno, M. (1990), “Auditing for performance evaluation”, The Accounting Review, Vol. 65 No. 3,
pp. 520-536.
Salamon, G. and Smith, E. (1979), “Corporate control and managerial misrepresentation of ?rm
performance”, The Bell Journal of Economics, Vol. 10 No. 1, pp. 319-328.
ARJ
26,2
164
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Schwarcz, S. (2002), “Enron and the use and abuse of special purpose entities in corporate
structure”, The Financier, Vol. 9, pp. 23-29.
Securities and Exchange Commission (2001), Securities Exchange Act of 1934, Release No. 44444,
Securities and Exchange Commission, available at: www.sec.gov/litigation/admin/
34-44444.htm
Sunder, S. (1997), Theory of Accounting and Control, South-Western, Cincinnati, OH.
Sunder, S. (2010), “Adverse effects of uniform written reporting standards on accounting practice,
education, and research”, Journal of Accounting & Public Policy, Vol. 29, pp. 99-114.
Tof?er, B. and Reingold, J. (2003), Final Accounting: Ambition, Greed, and the Fall of Arthur
Andersen, Broadway Books, New York, NY.
Watts, R. and Zimmerman, J. (1990), “Positive accounting theory: a ten year perspective”,
The Accounting Review, Vol. 65, pp. 31-156.
Further reading
Klein, A. (2002), “Audit committee, board of director characteristics, and earnings management”,
Journal of Accounting and Economics, Vol. 33, pp. 375-400.
Appendix 1. Illustrations 1-6
Below is the mathematical program that applies for Illustrations 1 through 6 where the report is
limited to g or b. No discretion is indicated by n ¼ 1, and discretion by n $ 2 (Table AI):
w
g
;w
b
min p
g
w
g
þ ð1 2p
g
Þw
b
subject to:
p
g
??????
w
g
p
þ ð1 2p
g
Þ
?????
w
b
p
2 c
H
$
U ðPÞ
p
g
??????
w
g
p
þ ð1 2p
g
Þ
?????
w
b
p
2 c
H
$ q
g
??????
w
g
p
þ ð1 2q
g
Þ
?????
w
b
p
ðICÞ
w
g
; w
b
; $ 0 ðBÞ
where n is the number of available signals, and:
p
g
¼ 1 2ð1 2pÞ
n
and q
g
¼ 1 2ð1 2qÞ
n
p ¼ probability of a signal of g, given H.
q ¼ probability of a signal of g, given L.
p
g
¼ probability manager reports g, given H.
q
g
¼ probability manager reports g, given L.
Illustration n p q p
g
q
g
c
H
U
¯
1 1 0.7 0.5 0.7 0.5 50 500
2 2 0.7 0.5 0.91 0.75 50 500
3 1 0.97 0.96 0.97 0.96 1 2,500
4 2 0.97 0.96 0.9991 0.9984 1 2,500
5 1 0.7 0.5 0.7 0.5 50 50
6 2 0.7 0.5 0.91 0.75 50 50
Table AI.
Observations
on accounting
discretion
165
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Appendix 2. Illustrations 7-8
Below are the programs that apply to Illustrations 7 and 8, in which the manager has been
granted discretion to choose from between two signals and the owner declares a preferred signal:
w
pg
;w
ng
;w
b
min p
pg
w
pg
þ p
ng
w
ng
þ p
b
w
b
subject to:
p
pg
???????
w
pg
p
þ p
ng
???????
w
ng
p
þ p
b
?????
w
b
p
2 c
H
$
U ðPÞ
p
pg
???????
w
pg
p
þ p
ng
???????
w
ng
p
þ p
b
?????
w
b
p
2 c
H
$ q
pg
???????
w
pg
p
þ q
ng
???????
w
ng
p
þ q
b
?????
w
b
p
ðICÞ
w
pg
; w
ng
w
b
; $ 0 ðBÞ
w
pg
¼ payment for a report of g from the preferred signal.
w
ng
¼ payment for a report of g from the non-preferred signal.
w
b
¼ payment for a report of b.
Probabilities given action H are: p
pg
¼ p, p
np
¼ 1 2ð1 2pÞ
2
2p, p
b
¼ 1 2½1 2ð1 2pÞ
2
?.
Probabilities given action L are: q
pg
¼ q q
np
¼ 1 2 (1 2 q)
2
2 q, q
b
¼ 1 2 [1 2 (1 2 q)
2
].
Illustration 7 (based on 3 and 4): p ¼ 0.97, q ¼ 0.96,U
¯
¼ 2,500, and c
H
¼ 1.
Illustration 8 (based on 5 and 6): p ¼ 0.7, q ¼ 0.5, U
¯
¼ 50, and c
H
¼ 50.
Corresponding author
Steven Schwartz can be contacted at: [email protected]
ARJ
26,2
166
To purchase reprints of this article please e-mail: [email protected]
Or visit our web site for further details: www.emeraldinsight.com/reprints
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
1
:
1
7
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_425370728.pdf