Operations Research Letters 36 (2008) 291–296

www.elsevier.com/locate/orl

The agency problem between the owner and the manager in real investment:
The bonus–audit relationship
Michi Nishihara a,∗ , Takashi Shibata b
a Center for the Study of Finance and Insurance, Osaka University, Osaka 560-8531, Japan
b Graduate School of Social Sciences, Tokyo Metropolitan University, Tokyo 192-0397, Japan

Received 13 January 2007; accepted 6 September 2007

Available online 23 December 2007

Abstract

This paper derives the owner’s optimal contract with a bonus-incentive and audit when the owner delegates the investment timing decision to a
manager with private information on an investment project. The optimal solution not only unifies the previous studies, but also accounts for actual
auditing systems in firms.
c 2007 Elsevier B.V. All rights reserved.

Keywords: Real options; Asymmetric information; Agency conflicts; Audit

1. Introduction
The real options approach has become an increasingly stan-
dard framework for the investment timing
decision in corporate finance (see, for example, [7]). Although
the early literature on real options (e.g., [6,12]) considered
the investment decision of a monopolist, more recent stud-
ies have investigated the problem of several firms competing
in the same market from a game theoretic approach (see [5]
for an overview). Grenadier [8] derived the equilibrium in-
vestment strategies of firms in a Cournot–Nash framework and
Weeds [15] incorporated equilibrium in a timing game in a real
options model. The effects of incomplete information about the
competitors were examined in [10,13].
While these studies have focused on the strategic interaction
with rival firms, Grenadier and Wang [9] investigated
investment timing in a decentralized firm where the owner
(principal) delegates the investment decision to the manager
(agent) who holds private information by combining the real
options approach and contract theory (for contract theory,
see the standard textbook [4]). In their model, asymmetric
information changes the investment behavior of the firm
∗ Corresponding author.
E-mail addresses: nishihara@sigmath.es.osaka-u.ac.jp (M. Nishihara),
tshibata@tmu.ac.jp (T. Shibata).
from the first-best no-agency case because the owner designs
the contract to provide a bonus-incentive for the manager
to truthfully reveal private information. Similar real options
models with agency conflicts have been also studied in [3,11].
Although these models consider only the carrot (i.e., giving
a bonus-incentive to the manager) as a measure for dealing
with agency conflicts, the owner can usually use not only the
carrot, but also the stick (i.e., auditing and fining the manager).
Naturally, the impact of auditing has been clarified in other
contexts (e.g., [1]).
In this paper, we incorporate an auditing mechanism into
a model following [9]. As far as the purpose of the paper is
concerned, we limit their original setup involving both hidden
information and action to just the case of hidden information.
We assume that the owner can utilize an auditing system that
fines the manager when a false report is detected, where the
higher the cost the owner pays in auditing, the greater the
probability of detection. We show that the optimal contract
is determined among three feasible types of contracts: the
bonus-incentive only mechanism, the joint bonus-incentive and
auditing mechanism, and the auditing only mechanism. This
is according to the relation between the auditing cost and the
amount of the penalty. Although a similar auditing technology
is introduced in [14], the findings in this paper are more
comprehensive and help connect previous results in this area.
Indeed, our solution includes the results of [9] in the bonus-

c 2007 Elsevier B.V. All rights reserved.

0167-6377/$ - see front matter
doi:10.1016/j.orl.2007.09.008
292 M. Nishihara, T. Shibata / Operations Research Letters 36 (2008) 291–296

incentive only region, the results of [14] in the joint bonus- where T is a set of all Ft stopping times, (Ft is the usual
incentive and auditing region, and the first-best no-agency filtration generated by z(t)), and r is a constant risk-free rate
solution as the limit of the auditing only region. satisfying r > α. In this paper, it is always assumed that
Furthermore, our results also give good account of the real the initial value P0 is sufficiently low that the firm has to
life relationship between audit and bonus-incentive as follows. wait for its exercise condition to be met. Using the standard
In the case where the manager may commit seriously outright method (see [7]), we obtain the value function W (P0 ; θi ) and
frauds such as embezzlement, the owner tries to prevent the the optimal stopping time τi∗ of problem (2) for θ = θi as
manager’s offense by using only audit, and stern punishment follows:
by the law improves the social welfare. On the other hand, in  β
P0
the case where the manager’s private benefit is not necessarily W (P0 ; θi ) = (Pi∗ + θi − K )
illegal, the owner is likely to prefer a bonus system such as Pi∗
stock options, and an owner’s too great demand may decrease τi∗ = inf{t ≥ 0 | P(t) ≥ Pi∗ } (3)
the social benefit. β
The paper is organized as follows. Section 2 provides Pi∗ = (K − θi ),
β −1
a brief review of [9]. Section 3 incorporates the auditing p
technology into the model and derives the owner’s optimal where β is defined by β = 1/2−µ/σ 2 + (µ/σ 2 − 1/2)2 + 2r/σ 2
contract after allowing for audit. The final section discusses (>1). The threshold Pi∗ is the optimal investment trigger for the
economic implications of our results involving both audit and owner who observes the value θi at time 0. Thus, the ex ante
the bonus-incentive. value of the owner’s option in the first-best no-agency setting
(denoted V ∗ (P0 )) becomes
2. A simple version of Grenadier and Wang [9]
V ∗ (P0 ) = q W (P0 ; θ1 ) + (1 − q)W (P0 ; θ2 )
This section provides a brief review of the results of [9].  β
P0
First, let us explain the setup. We consider a decentralized firm =q (P1∗ + θ1 − K )
P1∗
that faces the investment timing decision for a single project.
 β
We assume that the owner (principal) delegates the decision to P0
the manager (agent) and that both the owner and the manager + (1 − q) (P2∗ + θ2 − K ). (4)
P2∗
are risk-neutral. While both the owner and the manager know
the investment cost K , the value of the project consists of Now, let us turn to the principal–agent setting without
two components, namely the value P(t) that is observable auditing. In this case, the owner has the option to invest, but
to both the owner and the manager, and the value θ that is delegates the exercise decision to the manager. At time 0, the
privately observed only by the manager. Thus, the total value owner offers the manager a contract that commits the owner
of the project is P(t) + θ. For simplicity, we assume that the to pay the manager at the time of exercise. We assume no
observable value P(t) at time t obeys the following geometric opportunity for renegotiation exists. Although the commitment
Brownian motion: may lead to ex post inefficiency in investment timing, it
increases the ex ante value of the project. In fact, if the owner
dP(t) = α P(t)dt + σ P(t)dz(t) (t > 0), makes no contract with the manager, the owner’s ex ante option
P(0) = P0 , (1) value becomes (4) with q = 0. This is because the manager
hands the owner θ2 and makes θ1 − θ2 his/her own when the
where α ≥ 0, σ > 0 and P0 are given constants, and z(t) true value is θ1 . As discussed in [9], the optimal contract is
denotes the one-dimensional standard Brownian motion. included in a mechanism MGW = {(Pi , wi ) | i = 1, 2}
The private component θ potentially takes on two possible in which the owner pays the manager the bonus wi at the
values, θ1 or θ2 , with 0 ≤ θ2 < θ1 < K . We define 1θ = investment time when the manager exercises the investment
θ1 − θ2 . Before contracting, both the owner and the manager at time τi = inf{t ≥ 0 | P(t) ≥ Pi }. Here the superscript
know that the probability of drawing a higher quality project θ1 “GW” refers to the solution of [9]. Since the revelation principle
equals q. Immediately after making a contract with the owner (see [4]) ensures that the manager who observes θi faithfully
at time 0, the manager privately observes whether the project invests at the trigger Pi , the optimal contract is the solution of
is of a higher quality θ1 or a lower quality θ2 . Although the the problem of maximizing the owner’s ex ante option value of
manager’s one-time effort, which cannot be observed by the the investment:
owner, changes the likelihood q in [9], we exclude the effect  β
of the hidden action from the original setup. P0
max q (P1 + θ1 − K − w1 )
As a benchmark, we examine the case where there is no Pi ,wi P1
delegation of the exercise decision and the owner observes the  β
P0
true value of θ. In this case, the owner’s problem with given + (1 − q) (P2 + θ2 − K − w2 )
P2
θ = θi (i = 1, 2) becomes the following optimal stopping  β  β
problem: P0 P0
s.t. q w1 + (1 − q) w2 ≥ 0 (5)
P1 P2
W (P0 ; θi ) = sup E[e−r τi (P(τi ) + θi − K )], (2)
τi ∈T wi ≥ 0 (i = 1, 2)
 M. Nishihara, T. Shibata / Operations Research Letters 36 (2008) 291–296 293
 β  β
P0 P0 from the managerial post. This is because the manager does
w1 − (w2 + ∆θ ) ≥ 0
P1 P2 not in any case cheat the owner in the optimal contract in our
 β  β setting.
P0 P0
w2 − (w1 − ∆θ ) ≥ 0, Here, the cost function c(di ) and the penalty Γ are given
P2 P1
exogenously. We assume that the cost function c(di ) satisfies
where Pi > P0 . In the constraints of problem (5), the conditions c(0) = 0, limdi ↑1 c(di ) = +∞, c0 (di ) > 0 (di ∈
first and second inequalities correspond to the ex ante [0, 1)), and c00 (di ) > 0 (di ∈ [0, 1)). The first and
participation constraint and the ex post limited-liability third conditions are explicit from the property of auditing.
constraints, respectively, while the last two inequalities are The second assumption is realistically intuitive because no
the ex post incentive-compatibility constraints. The incentive- auditing can always detect the manager’s false report. The final
compatibility constraint means that with a truthful report, condition ensures the convexity of the cost function.
the manager who observes θ = θ1 (resp. θ = θ2 ) In this setting, the contract is designed as a mechanism
obtains the expected payoff (P0 /P1 )β w1 (resp. (P0 /P2 )β w2 ), MA = {(Pi , wi , di ) | i = 1, 2}, where the auditing level di for
which is larger than the expected payoff for a false report, the manager’s report θ = θi is added to the mechanism MGW .
(P0 /P2 )β (w2 + ∆θ ) (resp. (P0 /P1 )β (w1 − ∆θ )). Here the superscript “A” refers to the solution of the setting
In problem (5), it can be shown that the bonus w2 = 0 allowing audit. The optimal contract to the owner becomes the
and only the third inequality (i.e., the incentive-compatibility solution of the following problem:
condition for the manager who observes the better project value
θ1 ) binds. Then, the optimal solution {(PiGW , wiGW ) | i = 1, 2} P0 β
 
max q (P1 + θ1 − K − w1 − c(d1 ))
becomes Pi ,wi ,di P1
 !β   β
P ∗ P0
(P1GW , w1GW ) =  P1∗ , 1
∆θ  (6) + (1 − q) (P2 + θ2 − K − w2 − c(d2 ))
P2GW P2
 β  β
β P0 P0
   
q∆θ w1 + (1 − q) w2 ≥ 0
(P2GW , w2GW ) = K − θ2 + ,0 . (7) s.t. q
β −1 1−q P1 P2
(8)
For further details, see the solution for the hidden information wi ≥ 0 (i = 1, 2)
only region in [9]. It is worth noting that the trigger for
the higher quality project, P1GW , remains unchanged from the di ≥ 0 (i = 1, 2)
first-best trigger P1∗ defined by (3), while the trigger for the  β  β
P0 P0
lower quality project, P2GW , is larger than the first-best trigger w1 − (w2 + ∆θ − d2 Γ ) ≥ 0
P1 P2
P2∗ defined by (3). This is because the owner attempts to
decrease the information rent to the manager who observes  β  β
P0 P0
the higher quality project by deferring investment timing for w2 − (w1 − ∆θ − d1 Γ ) ≥ 0,
P2 P1
the lower quality project. The owner’s and manager’s ex ante
option values, πoGW and πmGW , respectively, are obtained by where Pi > P0 . The first two constraints of problem (8)
substituting the optimal solution {(PiGW , wiGW ) | i = 1, 2} are the same as those of problem (5), while the incentive-
into the objective function and the right-hand side of the first compatibility constraints of problem (8) include an additional
inequality of the constraints of problem (5). term, the expected penalty di Γ . It can be easily checked that
This is the result in the hidden information case of [9]. In the revelation principle holds in this case, as in the previous
the next section, we extend their analysis to a case allowing the setting, even if the penalty Γ is counted in the owner’s profit.
owner to audit the manager at a cost. Therefore, the owner does not make any contract to allow the
manager to report untruthfully.
3. Theoretical results
Proposition 1. The optimal contract {(PiA , wiA , diA ) | i =
This section derives the optimal contract involving the 1, 2} in the setting with auditing is given as follows:
bonus-incentive and audit. The owner detects the real value of Case A: 0 < Γ ≤ (1 − q)c0 (0)/q (bonus-incentive only region)
θ at probability di by paying the auditing cost c(di ) for the
manager’s report θ = θi when the manager executes the project. (P1A , w1A , d1A ) = (P1∗ , w1GW , 0)
We assume that the manager is fined the penalty Γ (>0) for (P2A , w2A , d2A ) = (P2GW , 0, 0),
cheating when a false report is detected. In general, as discussed
Case B: (1 − q)c0 (0)/q < Γ ≤ max{∆θ, (1 − q)c0 (∆θ/Γ )/q}
in [2], the society could suffer different damages according to
(joint bonus-incentive and auditing region)
the types of the punishment. However, we do not have to care
∗ β
 
whether the owner can receive the total amount of the fine, !
P
Γ from the manager, or whether a part of Γ represents the (P1A , w1A , d1A ) =  P1∗ , 1
(∆θ − d2A Γ ), 0
manager’s disutility from other punishment such as dismissal P2A
294 M. Nishihara, T. Shibata / Operations Research Letters 36 (2008) 291–296

β
 
(P2A , w2A , d2A ) = K − θ2
β −1

q
+ (∆θ − d2 Γ ) + c(d2 ) ,
A A
1−q
 
0 −1 qΓ
,


0, c
1−q
Case C: Γ > max{∆θ, (1 − q)c0 (∆θ/Γ )/q} (auditing only
region)
(P1A , w1A , d1A ) = (P1∗ , 0, 0)
(P2A , w2A , d2A )
β
 
= (K − θ2 + c(d2A )), 0, ∆θ/Γ .
β −1
Fig. 1. P2A , w1A and c(d2A ) for penalties Γ .
Proof. Note that in problem (8), the first constraint is induced
β
by the second constraints wi ≥ 0 (i = 1, 2), and P0 can be
ignored. We solve the problem (8) without the final constraint
(the incentive-compatibility constraint for the manager who
observes the bad value θ = θ2 ), and then also check that
the obtained solution satisfies the removed constraint. Let
{(PiA , wiA , diA ) | i = 1, 2} be the optimal solution of problem
(8) without the final constraint. It immediately follows that
w2A = 0, d1A = 0, and
−β −β
P1A w1A − P2A (∆θ − d2A Γ ) = 0. (9)
Let λi (i = 1, 2, 3) denote the Lagrangian multipliers
associated with the remaining constraints w1 ≥ 0, d2 ≥ 0, and
(9), respectively. That is, we form the Lagrangian
L(P1 , P2 , w1 , d2 ) = q P1 −β (P1 + θ1 − K − w1 ) Fig. 2. π2A (P0 ), πm
A (P ) and L A (P ) for penalties Γ .
0 0
+ (1 − q)P2 −β (P2 + θ2 − K − c(d2 )) + λ1 w1
the solution for Case B. If λ1 > 0 and λ2 = 0, from (9)–(13)
+ λ2 d2 + λ3 P1 −β w1 − P2 −β (∆θ − d2 Γ ) .


−β
we have the solution for Case C with λ1 = P1A (q − (1 −
The Karush–Kuhn–Tucker conditions are (9), q)c (d2 )/Γ ), λ3 = (1 − q)c (d2 )/Γ . If λ1 = 0 and λ2 > 0,
0 A 0 A
−β
∂L −β we obtain the solution for Case A with λ2 = P2A ((1 −
= q((−β + 1)P1A q)c0 (0) − qΓ ), λ3 = q. If λi > 0 (i = 1, 2), from (9) we
∂ P1
− β(θ1 − K − w1A )P1A
−β−1
) have ∆θ = 0, which contradicts ∆θ > 0. Taking into account
−β−1 that the conditions λi ≥ 0 (i = 1, 2, 3), d2A < 1 that result
− λ3 βw1A P1A = 0, (10) from limd2 ↑1 c(d2 ) = +∞, and the condition under which the
∂L −β solution of c0 (d2A ) = qΓ /(1 − q) exists, we can show that
= (1 − q)((−β + 1)P2A for a given Γ , the solution satisfying the Karush–Kuhn–Tucker
∂ P2
−β−1 conditions (9)–(14) is uniquely determined as the statement of
− β(θ2 − K − c(d2A ))P2A )
−β−1 Proposition 1. Furthermore, the solution explicitly satisfies the
+ λ3 β(∆θ − d2A Γ )P2A = 0, (11) final constraint in problem (8). 
∂L −β −β In Proposition 1, and as intuitively expected, the owner
= −q P1A + λ1 + λ3 P1A = 0, (12)
∂w1 neither gives any bonus for the manager’s bad report θ2 nor
∂L −β audits the manager’s good report θ1 . The investment trigger
= −(1 − q)c0 (d2A )P2A of the high quality project does not change from that of the
∂d2
−β no-agency setting as in the result of [9]. However, the other
+ λ2 + λ3 Γ P2A = 0, (13)
components of the contract and the owner’s strategy changes,
and depending on the auditing cost c( p) and the amount of the
penalty, Γ . Indeed, the contract is classified into three regions.
λ1 w1A = λ2 d2A = 0, λi ≥ 0 (i = 1, 2, 3). (14)
The solution changes from Case A to C via Case B as the
Let us now derive the solution of (9)–(14), depending on penalty Γ becomes larger, as observed in Figs. 1 and 2. In the
whether λi equals zero. If λ1 = λ2 = 0, we have λ3 = q and numerical example, we set the parameter values α = 0, r =
 M. Nishihara, T. Shibata / Operations Research Letters 36 (2008) 291–296
0.04, σ = 0.2, and K = 1 as in [7], and set P0 = 0.5, θ1 =
0.5, θ2 = 0, q = 0.5, and c(di ) = 0.5di /(1 − di ). Fig. 1
depicts the expected discounted cost of the bonus and auditing,
(P0 /P1A )β w1A , (P0 /P2A )β c(d2A ), together with the investment
trigger P2A . We do not illustrate quantities P1A , w2A , and d1A
because of P1A = P1∗ , w2A = d1A = 0.
Case A is likely to hold if the marginal cost of auditing is
high relative to the penalty Γ , or if the probability of drawing
the better project θ1 is low. In this case, the owner pays the
whole information rent to the manager without auditing, since
the auditing technology does not work at all. Conversely, in
Case C, where the penalty is severe and the auditing cost is not
so high, the owner uses only the auditing system without giving
any bonus to the manager. Case B is the intermediate case. In
this case, both the bonus-incentive and the audit are effective
by the trade-off between the auditing cost and the amount of
the penalty.
Let us explain the relation between Proposition 1 and the
results from previous studies. The solution in Case A coincides
with that of [9], i.e., (6) and (7). Then, it is readily seen that
P2∗ < P2A ≤ P2GW , πoGW (P0 ) ≤ πoA (P0 ) < V ∗ (P0 ), and
πmA (P0 ) ≤ πmGW (P0 ), where πoA and πmA denote the owner’s
and manager’s ex ante option values defined by
 β
P0
πoA (P0 ) = q (P1∗ + θ1 − K − w1A )
P1∗
!β
P0
+ (1 − q) (P2A + θ2 − K − c(d2A ))
P2A
 β
P0
πmA (P0 ) = q w1A .
P1∗
Note that πoA and πmA are exactly the same as πoGW and πmGW ,
respectively, for Case A. If we assume that Γ = ∆θ and
c0 (0) = 0, the solution is classified into Case B and agrees
with that of [14] where the owner can control the penalty Γ to a
limit ∆θ. Moreover, the solution in Case C converges to that of
the first-best no-agency case on letting Γ ↑ +∞. This appears
to correspond to Proposition 4 in [1].
We can see from Fig. 1 that P2A monotonically decreases to
P2 = 2 for penalties Γ . Fig. 2 indicates that the owner’s (resp.
∗
manager’s) ex ante option value πoA (resp. πmA ) monotonically
increases (resp. decreases) for Γ . These results can also
be easily proved, and therefore the proofs are omitted. In
particular, the monotone increase in the owner’s option value
with respect to the amount of the penalty is consistent with
the maximal punishment principle (see [1,14]). That is, the
owner imposes the maximum penalty if he/she can determine
the penalty within some limit.
We define the social loss L A by L A = V ∗ − (πoA + πmA ) as
in [9]. It should be noted that the social loss L A does not need to
involve the social cost of the punishment, which varies between
different kinds of punishment as mentioned in [2], since the
optimal contract precludes cheating. The social loss L A , unlike
the owner’s option value, is not necessarily monotonic for Γ
as observed in Fig. 2, although for sufficiently large Γ , it
295
monotonically decreases to zero (cf. Proposition 4 in [1]). This
means that a halfway penalty could only be of benefit to the
owner and brings with it inefficiency in terms of social surplus,
while a severe penalty improves social efficiency.
4. Economic insights
This section gives the motivation of considering both audit
and bonus-incentive in the contract described in the previous
section as well as the economic insights extracted from the
results that we obtained in the previous section. To do this, we
shall place our discussion in a real life context.
In practice, it is common for firms to employ independent
auditing systems in addition to their internal ones. There are
three types of independent audits, that is, financial statement
audit, operational audit, and compliance audit. We shall now
relate our results to the cases of financial statement audit and
operational audit.
By the means of a financial statement audit, the owner can
ensure the accuracy of the firm’s financial statements in its
business report. In other words, the owner attempts to prevent
the manager from making false statements regarding the firm’s
financial status and figures. When a manager’s outright frauds
such as window-dressing settlement and embezzlement are
detected by the audit, the manager is severely punished by
the relevant law (e.g., The Commercial Code, The Corporation
Law, The Securities and Exchange Law, The Penal Code, etc.
in the case of Japan). A well-known example is the WorldCom
scandal in which its managers were punished for its window-
dressing settlement which led to reported profits higher than the
actual profits. Another example is the Aramark case in which
its managers were charged with the embezzlement involving
underreporting of the company’s vending machine revenues.
For this kind of audit, the penalty Γ is expected to be much
larger than ∆θ, since Γ includes not only the compensation but
also further punishment by the law. Hence, we believe that Case
C in Proposition 1 is appropriate for this situation, even if the
auditing cost c is somewhat high. Here, the owner simply tries
to set the auditing level high enough to prevent the manager
from making false financial statements. Another logical finding
that we obtained under Case C is that severe penalty Γ brings
small social loss L A (see Fig. 2).
On the other hand, Cases A and B are more likely to
apply to operational audit than financial statement audit. The
purpose of operational audit is usually to assess the manager’s
efficiency, instead of detecting explicitly illegal frauds of the
manager. When the manager’s inefficient waste is pointed
out by the audit, the owner can force the manager to make
improvements by transferring a portion of the manager’s profit
to the owner. That is, in this case, Γ can be interpreted as the
owner’s punishment imposed on the manager rather than the
punishment by the law. The owner cannot claim Γ larger than
∆θ, because the manager’s act is not necessarily illegal in this
case. Accordingly, it is quite likely that the owner prefers the
bonus system to operational audit depending on the auditing
cost. The bonus system often takes the form of stock options in
real life. Note that in this situation the social welfare may be lost
296 M. Nishihara, T. Shibata / Operations Research Letters 36 (2008) 291–296
if the owner demands too much profit from the manager who
does not execute any illegal fraud. This intuition is captured
by our results for Case B where the social loss L A does not
necessarily show a monotone decrease with respect to Γ (see
Fig. 2).
Acknowledgments
The authors would like to thank an anonymous referee,
Professor Masao Fukushima, Professor Masaaki Kijima, and
Tony Wong for their helpful comments. This research was
partially supported by a MEXT Grant-in-Aid for Scientific
Research and the 21st Century COE Program “Informatics
Research Center for Development of Knowledge Society
Infrastructure”.
References
[1] D. Baron, D. Besanko, Regulation, asymmetric information, and auditing,
RAND Journal of Economics 15 (1984) 447–470.
[2] G. Becker, Crime and punishment: An economic approach, Journal of
Political Economy 76 (1985) 169–217.
[3] P. Bjerksund, G. Stensland, A self-enforced dynamic contract for
processing of natural resources, in: M. Brennan, L. Trigeorgis (Eds.),
Project Flexibility, Agency, and Competition, Oxford University Press,
Oxford, 2000, pp. 109–127.
[4] P. Bolton, M. Dewatripont, Contract Theory, MIT Press, London, 2005.
[5] M. Boyer, É. Gravel, P. Lasserre, Real options and strategic competition:
A survey, Working paper, Université de Montréal, 2004.
[6] A. Dixit, Entry and exit decisions under uncertainty, Journal of Political
Economy 97 (1989) 620–638.
[7] A. Dixit, R. Pindyck, Investment Under Uncertainty, Princeton University
Press, Princeton, 1994.
[8] S. Grenadier, Option exercise games: An application to the equilibrium
investment strategies of firms, Review of Financial Studies 15 (2002)
691–721.
[9] S. Grenadier, N. Wang, Investment timing, agency, and information,
Journal of Financial Economics 75 (2005) 493–533.
[10] B. Lambrecht, W. Perraudin, Real options and preemption under
incomplete information, Journal of Economic Dynamics and Control 27
(2003) 619–643.
[11] J. Mæland, Valuation of an irreversible investment: Private information
about a stochastic investment cost, Working paper, Norwegian School of
Economics and Business Administration, 2005.
[12] R. McDonald, D. Siegel, The value of waiting to invest, Quarterly Journal
of Economics 101 (1986) 707–727.
[13] M. Nishihara, M. Fukushima, Evaluation of firm’s loss due to incomplete
information in real investment decision, European Journal of Operational
Research 188 (2008) 569–585.
[14] T. Shibata, Investment timing, asymmetric information, and audit
structure: A real options framework, Working paper, Kyoto University,
2006.
[15] H. Weeds, Strategic delay in a real options model of R&D competition,
Review of Economic Studies 69 (2002) 729–747.