Incentives and Investment Timing: Real Options in a
Principal-Agent Setting∗
Steven R. Grenadier† Neng Wang‡
January 2003
Abstract
In the standard real options approach to investment, the owner of the optiondecides when to exercise the investment option. However, in many real-worldsettings, investment decisions are delegated to managers. This article providesa model of optimal contracting in a continuous-time principal-agent setting inwhich there is both moral hazard and adverse selection. We show that theunderlying option can be decomposed into two components: a manager’s optionand an owner’s option. The specification of the manager’s option is determinedby a compensation contract, and must provide an incentive for the managerto both extend effort and to exercise optimally. The residual option payoutgoes to the owner. The implied investment behavior differs significantly fromthat of the first-best no-agency solution. In particular, there will be greaterinertia in investment, as the model leads to the manager having an even greater“option to wait” than the owner. The interplay between the twin forces of hiddeninformation and hidden action leads to markedly different investment outcomesthan when only one of the two forces is at work.
∗We thank Harrison Hong, David Scharfstein, Jeff Zwiebel, and especially Erwan Morellec, forhelpful comments.
†Graduate School of Business, Stanford University and NBER. Email: [email protected].‡William E. Simon School of Business Administration, University of Rochester. Email:
1
1 Introduction
The real options framework has proven to be a powerful and flexible approach for
analyzing investment under uncertainty. Essentially, the real options approach posits
that the opportunity to invest in a project is analogous to an American call option on
the investment project. Once that analogy is made, the vast and rigorous machinery
of financial options theory is at the disposal of real investment analysis. Thus, the
timing of investment is economically equivalent to the optimal exercise decision for an
option. The real options approach is well summarized in Dixit and Pindyck (1994) and
Trigeorgis (1996).1
In the standard real options paradigm, there are no agency conflicts as it assumed
that the option’s owner makes the exercise decision. Thus, there is a stark assumption
that there is no delegation of investment decisions from owners to managers, and
hence no agency frictions that can impede the optimal timing of investment. A large
economics literature extensively studies principal-agent conflicts.2 In this paper, we
extend the real options framework to the principal-agent setting, where the manager is
delegated the decision as to when to invest. In this setting, managers can influence the
quality of the underlying option through their efforts, and where the information about
project quality is known only by the manager prior to investment. In the presence
of both hidden action and hidden information, we formulate an optimal contract that
provides an option value that is as close as possible to the first-best setting in which
no such agency conflicts exist.
In order to highlight the importance of including principal-agent considerations into
real options models, consider the following two examples of well-known applications.
Example 1: The US Government’s Delegation of Offshore Oil Exploration to Private
Firms
1The application of the real options approach to investment is quite broad. Brennan and Schwartz(1985) use an option pricing approach to analyze investment in natural resources. McDonald andSiegel (1986) provided the standard continuous-time framework for analysis of a firm’s investment ina single project. Majd and Pindyck (1987) enrich the analysis with a time-to-build feature. Dixit(1989) uses the real option approach to examine entry and exit from a productive activity. Triantisand Hodder (1990) analyze manufacturing flexibility as an option. Titman (1985) and Williams (1991)use the real options approach to analyze real estate development.
2See Mirrlees (1976) and Holmstrom (1979), for early formal models on principal-agent relationship.
2
A prototypical case of a real option is oil exploration. Paddock, Siegel, and Smith
(1988) model the option-like characteristics of offshore oil leases. Using an option
analogy, the right to drill for oil represents an American call option on the market value
of the oil reserves with an exercise price equal to the cost of extraction. In the case of
offshore oil exploration, there exists a clear principal-agent problem in that the U.S.
government delegates (via auctions) the production rights to private firms.3 Thus, the
principal (government) must use a contract to motivate the agent (private oil firm) to
exercise their option as efficiently as possible.4 Clearly, both moral hazard and adverse
selection problems may exist. The amount of research, effort and technology used by
the firm are not easily observed, yet can have a material impact on the ultimate costs
and efficiency of exploration. In addition, the amount and nature of the underlying oil
deposits will be known with greater certainty by the firm than the government. The
actual contract offered by the government entails that an annual royalty be paid based
on value of extracted oil, along with a fixed annual fee.
Example 2: Agency and the Management of Commercial Real Estate
Another example of a real option is the leasing of vacant space. As presented in
Grenadier (1995), the leasing of space is analogous to an American call option where the
underlying asset is the present value of lease cash flows and the exercise price equals the
lump sum costs of leasing such as tenant improvement costs. In the U.S., corporations
and institutional investors own approximately two thirds of nonresidential real estate
(Institute of Real Estate Management, 1991).5 As pointed out by Williams (2001), the
agency problems associated with owning and managing certain types of commercial
real estate can be severe. In particular, the leasing of vacant space in older, less-than-
pristine commercial buildings is prone to moral hazard and adverse selection problems.
Leasing costs such as finding tenants (or monitoring a leasing agent), negotiating with
tenants, and fixing up space to suit new tenants are often private, and not easily
verifiable or contractible. The owner of the building must provide an incentive for the
3See Hendricks, Porter and Boudreau (1987) and Hendricks and Kovenock (1989) for more detailsabout such offshore oil leases.
4While other objectives may be possible, we assume that the government’s objective is to achievethe greatest value from the potential oil deposit.
5Miles and Tolleson (1997) provide a conservative estimate of the value of U.S. commercial realestate of $4 trillion, as of 1997.
3
manager to supply the appropriate amount of effort in order to make payoff from leasing
to new tenants as close to the first-best level as possible. A compensation contract
must be constructed that is based on the observable value of the rent achieved by the
manager. For example, leasing agents (delegated the role of finding new tenants) are
typically paid a percentage of the rent obtained, with leasing bonuses and reputation
effects adding some non-linear compensation features. In addition, property managers
(delegated the role of maintaining a property) are compensated in a similar manner.
In the model, an owner delegates the option exercise decision to a manager. Thus,
the timing of investment is determined by the manager. The true quality of the project
can be high or low. However, through effort, the manager can influence the likelihood
that the quality of the project is high. This unobserved effort represents the hidden
action component of the model. In addition, while the manager learns the true qual-
ity of the project after the effort decision is made, this information is not revealed to
the owner. This represents the hidden information component of the model. A com-
pensation contract must be contingent on the observable realized value of the project
upon exercise. This contract must provide two types of incentives to the manager:
an incentive to extend effort and an incentive to exercise as close as possible to the
value-maximizing stopping time. Importantly, we show that the underlying option can
be decomposed into two components: a “manager’s option” and an “owner’s option.”
The manager’s option has a payout upon exercise that is a function of the compensation
contract. Based on this contractual payout, the manager determines the exercise time.
The owner’s option has a payout, received at the manager’s chosen exercise time, equal
to the payoff from the underlying option minus the manager’s compensation. The
model provides the solution for the optimal compensation contract that comes as close
as possible to the first-best no-agency solution.
The model implies investment behavior substantially different from that of the stan-
dard real options approach with no agency problems. In general, managers will display
greater inertia in their investment behavior, as they will invest later than implied by
the first-best solution. In essence, this is a result of the manager (even in an optimal
contract) not having a full ownership stake in the option payoff. This less than full
ownership interest implies that the manager has a greater “option to wait” than the
4
owner. Such investment delay due to incentive problems is consistent with the re-
sults in Bertrand and Mullainathan (2000), where they look at the compensation and
investment behavior of manufacturing firms following the passage of anti-takeover legis-
lation. Presumably, following such legislation, managers become more entrenched and
the divergence of interests between owners and managers becomes more pronounced.
They find that managers display more inertia in their investment policy following the
anti-takeover legislation. This is consistent with our model, and not as consistent with
the view that less disciplined managers behave like empire builders.
An important aspect of the model is the interaction of hidden action and hidden
information. In fact, we find that the nature of the optimal contract depends explicitly
on the relative importance of these two forces. While we focus on the most interesting
case in which both forces play a role in the optimal contract, it is instructive to consider
two extremes. If the cost-benefit ratio of effort (a measure of the strength of the
hidden effort component) is very low, then the hidden action component disappears
from the optimal contract terms. Thus, if the nature of the underlying option is
such that inducing effort is sufficiently inexpensive, then we are left with a simple
problem of hidden information, and the contract will simply reward the manager with
informational rent. This is the setting of Maeland (2002), which considers a real options
setting with only hidden information about the costs of investment. Conversely, if the
cost-benefit ratio of effort is very high, then the hidden action component dominates
the optimal contract. The cost of inducing effort is so high as to no longer necessitate
the payment of informational rents.
It is only when both forces are in effect that the optimal compensation contract must
trade off the relative costs and benefits of inducing effort and information revelation.
Interestingly, the interplay between hidden information and hidden action may actually
improve the efficiency in investment timing decision. This is because the compensation
that is needed to induce effort provides the manager with a greater ownership stake in
the underlying option, making the manager’s timing incentives closer to those of the
first-best. Bernardo, Cai and Luo (2001) analyze capital budgeting and compensation
in an investment context with moral hazard and adverse selection. However, their
focus in on the amount of investment rather than the timing of investment.
5
In a later section, we generalize the model to allow for managers to display greater
impatience than owners. For example, managers may have a shorter horizon (due
to job loss, death, etc.), or may appear to use a higher discount rate due to being
less diversified or liquidity constrained. This generalization leads to very different
predictions about investment timing. While the basic model predicts that investment
will occur later than the first-best case, in this generalized setting investment can
actually occur earlier than the first-best case. The intuition is that the marginal
benefit to the manager of receiving compensation at an earlier point in time is less
than the marginal cost to the owner.
The remainder of the paper is organized as follows. Section 2 describes the setup
of the model and derives the simplified optimization program. Section 3 solves for the
optimal contracts in three separate regions. In Section 4 we analyze the implications on
investment lags and the erosion of the option value due to the agency problem. Section
5 shows that both under- and over-investment problems may arise, when the manager
and the owner have different degrees of impatience. Section 6 concludes. Appendices
contain the solution details of the optimal contracts.
2 Model
In this section, we begin with a description of the model. We then, as a useful bench-
mark, provide the solution to the first-best no-agency investment problem. Finally, we
present and simplify the full principal-agent optimization problem faced by the owner.
2.1 Setup
The principal owns an option to invest in a single project. Let the value P (t) of the
underlying project evolve as a geometric Brownian motion:
dP (t) = αP (t) dt+ σP (t) dZ(t), (2.1)
where α is the instantaneous conditional expected percentage change in P (t) per unit
time, σ is the instantaneous conditional standard deviation per unit time, and dZ is
the increment of a standard Wiener process.
6
We assume that the principal (owner) delegates the exercise decision to an agent
(manager). This desire to delegate investment authority is premised on the fact that it
is more costly for the owner to exercise the option than it is for the manager, perhaps
due to the manager’s possessing some specialized skill, or to a higher opportunity cost
for the owner. Both the owner and the manager are risk neutral,6 with the risk-free
rate of interest denoted by r. For convergence, we assume that r > α. The value of
the project, P (t), is observable to both the owner and manager at all times. Let P
equal the current value of the project, P = P (0).
The exercise cost may take on two possible values: θ1 or θ2, with θ1 < θ2, determined
randomly. One may interpret a draw of θ1 as a “higher quality” project and a draw
of θ2 as a ”lower quality” project. However, the manager may affect the likelihood of
drawing θ1 by exerting a one-time effort, at time 0. If the manager exerts no effort (or
a low amount of effort with a normalized cost of zero), the probability of drawing a
low exercise cost θ1 equals qL. However, if the manager exerts effort, he incurs a cost
ξ > 0 at time 0, but increases the likelihood of drawing a low exercise cost θ1 from qL
to qH . Immediately after his exerting effort, the manager observes his realized exercise
cost, and knows if the project is of higher or lower quality.
The owner does not observe the realized exercise cost, but observes and can contract
on the trigger level at which the manager exercises the option (the value of the project
upon investment). The manager is paid immediately after he exercises the option.
The manager is always free to walk away and not exercise the option. That is, he has
limited liability, whether his exercise cost is low or high.7
In summary, the owner faces a problem with both hidden information (the owner
does not observe the true realization of θ) and hidden action (the owner cannot verify
the manager’s effort level). Before analyzing the optimal contracting, we first briefly
review the first-best no-agency solution that is used as the benchmark.
6We rule out the time-0 selling-the-firm contract between the owner and the manager. This maybe justified if the manager is liquidity constrained and cannot, for example, borrow from a bank.
7The limited-liability condition is essential in delivering the investment inefficiency result in thiscontext. Otherwise, with risk-neutrality assumptions for both the owner and the manager, and no lim-ited liability, the first-best optimal investment timing may be achieved even in the presence of hiddeninformation and hidden action [Innes (1990)]. An alternative mechanism for generating investmentinefficiency in an agency context is to assume risk aversion.
7
2.2 First-Best Benchmark (The Standard Real Options Case)
When the exercise cost θ is observed and contractible, then the first-best timing of
investment may be made. Let W (P ; θ) denote the value of the owner’s option, in a
world where θ is a known parameter. Using standard arguments [i.e. Dixit and Pindyck
(1994)], W (P ; θ) must solve the following differential equation:
0 =1
2σ2P 2WPP + αPWP − rW. (2.2)
Differential equation (2.2) must be solved subject to appropriate boundary condi-
tions. These boundary conditions serve to ensure that an optimal exercise strategy is
chosen:
W [P ∗(θ), θ] = P ∗(θ) − θ, (2.3)
WP [P ∗(θ), θ] = 1, (2.4)
limP→0
W (P, θ) = 0. (2.5)
Here, P ∗(θ) is the value of P (t) that triggers entry. The first boundary condition is the
value-matching condition. It simply states that at the moment the option is exercised,
the expected payoff is P ∗(θ)−θ. The second boundary condition is the smooth-pasting
or high-contact condition.8 This condition ensures that the exercise trigger is chosen
so as to maximize the value of the option. The third boundary condition reflects the
fact that zero is an absorbing barrier for P (t).
Closed-form solutions forW (P, θ) and the exercise trigger P ∗(θ) are easily obtained.
The value of the first-best option value and exercise trigger can be written as:
W (P ; θ) =
(
PP ∗(θ)
)β
[P ∗(θ) − θ] , for P < P ∗(θ),
P − θ, for P ≥ P ∗(θ),
(2.6)
where
P ∗(θ) =β
β − 1θ, (2.7)
and
β =1
σ2
−(α− σ2
2
)+
√(α− σ2
2
)2
+ 2rσ2
> 1. (2.8)
8See Merton (1973) for a discussion of the high-contact condition.
8
Note that P ∗′(θ) < 0, P ∗′′(θ) ≥ 0.
Since the realized value of θ can be either θ1 or θ2, we denote P ∗(θ1) = P ∗1 and
P ∗(θ2) = P ∗2 . We shall always assume the current value of the project, P , is less
than P ∗1 . The ex-ante value of the owner’s option under high effort and no agency is
qHW (P ; θ1) + (1 − qH)W (P ; θ2). We can write this first-best option value, V ∗(P ), as:
V ∗(P ) = qH
(P
P ∗1
)βθ1
β − 1+ (1 − qH)
(P
P ∗2
)βθ2
β − 1. (2.9)
It will prove useful in future calculations to define the present value of one dollar
received at the first moment that trigger P is reached, for i = 1, 2. Denote this present
value operator as D(P ; P ). This is simply the solution to differential equation (2.2)
subject to the boundary conditions that D(P ; P ) = 1, and D(0; P ) = 0. The solution
can be written as:
D(P ; P ) =
(P
P
)β
, P ≤ P . (2.10)
2.3 A Principal-Agent Setting
The owner offers the manager a contract at time zero. The contract specifies a payment
made to the manager, paid at the time of exercise. The owner is committed to
implementing the contract. The payment can be made contingent on the realized value
of the project at the time of exercise. Thus, in principle, for any value of P received
by the owner at the time of exercise, a contracted wage, w(P ), can be specified.
The principal-agent setting leads to a decomposition of the underlying option into
two options: an owner’s option and a manager’s option. The owner’s option has a payoff
function of P − w(P ), and the manager’s option has a payoff function of w(P ) − θ.
Obviously, the sum of these payoff functions equals the payoff of the underlying option.
The manager’s option is of the tradition American call option variety, since the manager
chooses the exercise time to maximize the value of his option. However, in this
optimal contracting setting, it is the owner that ultimately controls the timing of
exercise through the choice of contract parameters that induce the exercise policy that
maximizes the value of its option. In addition, the manager also possesses a compound
option, since the manager has the option to exert effort at time zero. The properties
of the manager’s option are thus contingent upon this initial effort choice.
9
Note that the problem has been framed as if the agent will pay the exercise cost of
θ. However, it is observationally equivalent to have the owner to pay the exercise cost
θ, if we redefine the net payment to the agent as s(P ) = w(P ) − θ. Then, the owner’s
payoff is P − s(P ) − θ and the manager’s payoff is s(P ). Both formulations provide
the same result. Throughout this paper, we will use the wage-value pair (w,P ) as the
contracting instruments.
Since there are only two possible values of θ, for any w(P ) schedule, there can be
at most two wage/value pairs that will be chosen by the manager.9 Thus, the contract
need only include two wage/value pairs from which the manager can choose: one that
will be chosen by a manager with an exercise cost of θ1, and one chosen by a manager
with an exercise cost of θ2. Therefore, the owner will offer a contract that promises a
wage of w1 if the option is exercised at P1 and a wage of w2 if the option is exercised
at P2. The revelation principle will ensure that a manager with an exercise cost of θ1
will exercise at the P1 trigger, and a manager with an exercise cost of θ2 will exercise
at the P2 trigger.
The owner’s option has a payout function of P1 − w1, if θ = θ1, and P2 − w2, if
θ = θ2. Thus, using function D( · ; · ) derived in (2.10), conditional on the manager
exerting effort, the value of the owner’s option, πo(P ;w1, w2, P1, P2), can be written as:
πo(P ;w1, w2, P1, P2) = qHD(P ;P1)(P1 − w1) + (1 − qH)D(P ;P2)(P2 − w2). (2.11)
The manager’s option has a payout function of w1−θ1, if θ = θ1, and w2−θ2, if θ = θ2.
Similarly, conditional on the manager exerting effort, the value of the manager’s option,
πm(P ;w1, w2, P1, P2), can be written as:
πm(P ;w1, w2, P1, P2) = qH D(P ;P1)(w1 − θ1) + (1 − qH)D(P ;P2)(w2 − θ2). (2.12)
For notational simplicity, we will sometimes drop the parameter arguments and simply
write the owner’s and manager’s option values as πo(P ), and πm(P ), respectively.
The owner’s objective is to maximize its option value through its choice of the
contract terms w1, w2, P1, and P2. Thus, the owner solves the following optimization
9We allow for the possibility of a pooling equilibrium in which only one wage/value pair is offered.However, this pooling equilibrium will always be dominated by a separating equilibrium with twowage/value pairs.
10
problem:
maxw1, w2, P1, P2
qH
(P
P1
)β
(P1 − w1) + (1 − qH)
(P
P2
)β
(P2 − w2) . (2.13)
This optimization is subject to a variety of constraints induced by the hidden infor-
mation and hidden action of the manager. The contract must induce the manager to
accept the contract, exert effort, exercise at the trigger P1 if θ = θ1, and exercise at
the trigger P2 if θ = θ2. It is the specification of these constraints to which we now
turn.
There are both ex-ante and ex-post constraints involving the manager’s option
value, πm( · ), derived in (2.12). The ex-ante constraints ensure that the manager
exerts effort and that the contract is accepted. These are the traditional constraints in
a static moral hazard setting.
• ex-ante incentive constraint:
qH
(P
P1
)β
(w1 − θ1) + (1 − qH)
(P
P2
)β
(w2 − θ2) − ξ ≥
qL
(P
P1
)β
(w1 − θ1) + (1 − qL)
(P
P2
)β
(w2 − θ2). (2.14)
The left side of this inequality is the value of the manager’s option if effort is exerted
minus the cost of effort. The right side is the value of the manager’s option if no effort
is exerted. This constraint ensures that the manager will exert effort.
• ex-ante participation constraint:
qH
(P
P1
)β
(w1 − θ1) + (1 − qH)
(P
P2
)β
(w2 − θ2) − ξ ≥ 0. (2.15)
This constraint ensures that the total value to the manager of accepting the contract
is non-negative.
The ex-post constraints ensure that the manager, in forming its optimal exercise
policy, acts in accordance with the owner’s beliefs. These are in keeping with the
revelation principle, in that the θ1-manager will exercise at P1 and the θ2-manager will
exercise at P2.
11
• ex-post incentive constraints (IC):
w1 − θ1 ≥(P1
P2
)β
(w2 − θ1) , (2.16)
w1 − θ2 ≤(P1
P2
)β
(w2 − θ2) . (2.17)
The first inequality ensures that a manager with exercise cost θ1 does indeed choose
to exercise at the trigger P1. This is because the value of the payout at trigger P1 is
seen to be at least as great as the present value of the payout obtained by waiting until
trigger P2 is reached.10 The second inequality ensures that a manager with exercise
cost θ2 does indeed choose to exercise at the trigger P2. This is because the value of
the payout at trigger P1 is seen to be less than or equal to the present value of the
payout obtained by waiting until trigger P2 is reached. Thus, these ex-post constraints
ensure that the revelation principle holds.
• ex-post limited-liability constraints:
wi − θi ≥ 0, i = 1, 2. (2.18)
The manager is thus never forced to lose any money by participating in the deal, ex-ante
or ex-post.
Therefore, the owner’s problem has a total of six inequality constraints: the ex-
ante incentive and participation constraints, and each of the two ex-post incentive and
limited-liability constraints. Fortunately, the following three Lemmas simplify the
problem in that we can reduce the number of constraints to three.
Lemma 1. The limited-liability condition for type-θ1 manager does not bind. That is,
w1 > θ1.
Proof 1. (P
P1
)β
(w1 − θ1) ≥(P
P2
)β
(w2 − θ1) ≥(P
P2
)β
∆θ > 0,
where ∆θ = θ2 − θ1 > 0. The first and second inequality follow from (2.16) and (2.18),
respectively.
10This argument is contingent upon P1 ≤ P2. We show below that this is always the case.
12
Re-arranging the IC constraint (2.14) gives
(P
P1
)β
(w1 − θ1) −(P
P2
)β
(w2 − θ2) ≥ ξ
∆q, (2.19)
where ∆q = qH − qL. The left-hand side of (2.19) is the expected excess surplus for a
type-θ1 manager over the surplus for a type-θ2 manager. The right-hand side measures
cost of exerting high effort, normalized by ∆q, the incremental change of the likelihood
of drawing low exercising cost θ1. Inequality (2.19) states that the manager will exert
high effort, if the incremental benefit from doing so is high enough.
In order to motivate the manager to exert high effort, we need to reward the manager
with a surplus larger than zero, which is the manager’s reservation value. This leads
to the following result.
Lemma 2. The ex-ante participation constraint (2.15) does not bind.
Proof 2.(P
P1
)β
(w1 − θ1) +1 − qH
qH
(P
P2
)β
(w2 − θ2) − ξ
qH
≥ ξ
∆q− ξ
qH
> 0,
where the first inequality follows from ex-ante IC constraint (2.19) and the limited
liability condition for type-θ2 manager.
The following result states that there is no rent (not counting time-0 sunk effort
cost) for the type-θ2 manager.
Lemma 3. The limited liability for type-θ2 manager binds, in that w2 = θ2.
The intuition is straightforward. Giving type-θ2 manager any positive rent ex-
post implies a higher rent for type-θ1 manager in order to meet the type-θ1 manager’s
IC constraint. In order to minimize the rents subject to the manager’s IR and IC
constraints, the owner shall give type-θ2 manager zero ex-post rent: w2 = θ2. A
formal proof of this Lemma appears in Appendix ??.
Together, Lemmas 1, 2, and 3 simplify the owner’s optimization problem as follows:
maxw1, P1, P2
qH
(P
P1
)β
(P1 − w1) + (1 − qH)
(P
P2
)β
(P2 − θ2), (2.20)
13
subject to
(P
P1
)β
(w1 − θ1) ≥(P
P2
)β
(θ2 − θ1) (2.21)
0 ≥(P
P1
)β
(w1 − θ2) (2.22)
(P
P1
)β
(w1 − θ1) ≥ ξ
∆q. (2.23)
At least one of (2.21) and (2.23) binds. Otherwise, reducing w1 will increase the
owner’s value strictly, without violating constraint (2.22). Based on which of these two
constraints binds, the resulting contract will differ.
In the next section we explicitly solve this (now simplified) constrained optimization.
We will find that there exist three equilibrium solutions to the principal-agent problem,
the particular solution depending on which of three disjoint regions a function of the
initial parameters resides.
3 Model Solution: Optimal Contracts
In this section, we provide the solution to the optimal contracting problem described
in the previous section: maximizing (2.20) subject to inequality constraints (2.21),
(2.22), and (2.23). We find that the nature of the solution depends on the initial
parameter values. In particular, the solution depends explicitly on the magnitude
of the cost-benefit ratio of the manager’s effort. Depending on this magnitude, the
optimal contract can take on three possible types: a “joint hidden information/ hidden
action” type, a “pure hidden information” type, and a “pure hidden action” type.
We first define the three regions that determine the nature of the optimal contract.
The key to the contract is the magnitude of the ratio of costs to benefits of the manager’s
effort, defined by ξ/∆q. The numerator is the direct cost of extending effort, and the
denominator is the change in the likelihood of achieving a low exercise cost due to
effort. The regions are then defined by where this cost-benefit ratio falls relative to the
present value of receiving a cash flow at three particular trigger values: P ∗1 = P ∗(θ1),
P ∗2 = P ∗(θ2), and KH = P ∗(ψH), where ψH = θ2 + qH∆θ/(1 − qH) > θ2. Using the
present value operator D( · ; · ), these present values can be expressed as D(P ; P ∗1 )∆θ,
14
D(P ; P ∗2 )∆θ, and D(P ; KH)∆θ, respectively. These present values are ordered by
D(P ; KH)∆θ < D(P ; P ∗2 )∆θ < D(P ; KH)∆θ. We now define the three critical
regions by whether the cost-benefit ratio of extending effort is medium, low or high.
1 The Joint Hidden-Information/Hidden-Action Region (medium cost-benefit
ratio): (P
KH
)β
∆θ ≤ ξ
∆q≤(P
P ∗2
)β
∆θ. (3.24)
2 The Hidden-Information Region (low cost-benefit ratio):
ξ
∆q<
(P
KH
)β
∆θ. (3.25)
3 The Hidden-Action Region (high cost-benefit ratio):
(P
P ∗2
)β
∆θ <ξ
∆q<
(P
P ∗1
)β
∆θ. (3.26)
Note that another potential region in which ξ/∆q > D(P ; P ∗1 )∆θ = ∆θ exists,
however in this range the costs of effort are so high as to no longer justify the exertion
of effort in equilibrium. Thus, we do not consider this region.11
We now describe the optimal contract in each of the three regions. Each contract is
specified by two wage/trigger-value pairs: (w1, P1), and (w2, P2). The proofs detailing
the solution are provided in Appendix A.
3.1 Joint Hidden Information/Hidden Action Region
In the parameter range defined by D(P ;KH)∆θ ≤ ξ/∆q ≤ D(P ; P ∗2 )∆θ, the optimal
contract can be written as:
P1 = P ∗1 , (3.27)
P2 = K2 = P
(∆q∆θ
ξ
)1/β
, (3.28)
w1 = θ1 +
(P ∗
1
K2
)β
∆θ, (3.29)
w2 = θ2. (3.30)
11A proof of this result is available from the authors by request.
15
Note that in this range, P2 > P ∗2 , and thus a θ2-type manager delays investment
beyond the first-best full information trigger. In addition, while the θ2-type manager
receives no net compensation (i.e., w2 = θ2), the θ1-type manager receives positive
information rents (i.e., w1 > θ1).
Since the manager must be induced into providing effort, w1 must be high enough
to provide enough compensation for the ex-ante IC constraint (2.23) to bind. This
reflects the “hidden action” component of the contract. In addition, given the surplus
wage paid to the θ1-type manager, the exercise trigger P2 must be high enough to
dissuade the θ1-type manager from deviating from the equilibrium first-best trigger
P ∗1 . Thus, in this region P2 is set so that the ex-post IC constraint (2.21) binds,
ensuring that the revelation principle holds. This requires that P2 be above the full-
information trigger P ∗2 . This deviation from the full-information trigger reflects the
“hidden information” component of the contract.
We can use these contract terms to place a value on the owner’s and manager’s op-
tion values. The owner’s and manager’s option values, πo(P ) and πm(P ), respectively,
can be written as:
πo(P ) = qH
(P
P ∗1
)βθ1
β − 1+ (1 − qH)
(P
K2
)β
(K2 − θ2) − qHξ
∆q, (3.31)
πm(P ) = qHξ
∆q. (3.32)
The owner’s option deviates from the first-best value, V ∗(P ) in (2.9), for two reasons.
First, the exercise trigger in the high exercise cost state is equal to K2 rather than P ∗2 ;
this is seen in the second term of the expression for πo(P ). Second, the manager must
be compensated for its effort through the wage paid in the low exercise cost state; this
is evidenced in the third term of the expression for πo(P ). The manager’s option value
is simply equal to the present value of the surplus he receives should he be a θ1-type
manager.
16
3.2 Hidden Information Only Region
In the parameter range, defined by ξ/∆q < D(P ; KH)∆θ, the optimal contract can
be written as:
P1 = P ∗1 , (3.33)
P2 = KH = P ∗(ψH), (3.34)
w1 = θ1 +
(P ∗
1
KH
)β
∆θ, (3.35)
w2 = θ2, (3.36)
where we recall that ψH = θ2 + qH∆θ/(1 − qH) > θ2.
In this region, the net costs of effort are low enough so that there is no need for
the firm to have to compensate the manager for extending effort. In this range, the
ex-ante IC constraint does not bind, and therefore the cost of effort does not find
its way into the optimal contract.12 The compensation the θ1-type manager receives
is purely an information rent that induces the manager to exercise at the first-best
trigger P ∗1 , in accordance with the revelation principle. Since w1 no longer needs to
be increased in order to induce effort (as in the joint region previously discussed), the
P2 trigger needs to be higher (relative to the first-best trigger P ∗2 ) in order to dissuade
the θ1-type manager from deviating from the equilibrium first-best trigger P ∗1 . That
is, in this region, P2 is higher than it was in the joint region. Therefore, perhaps
surprisingly, moral hazard serves to increase investment efficiency since the increased
share of the firm that must go to compensate the manager leads the manager to more
fully internalize the benefits of efficient investment timing.
We can use these contract terms to place a value on the owner’s and manager’s op-
tion values. The owner’s and manager’s option values, πo(P ) and πm(P ), respectively,
can be written as:
πo(P ) = qH
(P
P ∗1
)βθ1
β − 1+ (1 − qH)
(P
KH
)βψH
β − 1, (3.37)
πm(P ) = qH
(P
KH
)β
∆θ. (3.38)
12In this region, where there is no moral hazard component, the contract takes essentially the sameform as that in Maeland (2002).
17
It is interesting to note that the owner’s option value is observationally equivalent to
the first-best solution V ∗(P ) characterized in (2.7), when one substitutes the higher
exercise cost ψH for θ2. Thus, the impact of the costs of hidden information is fully
embodied by an increase in the cost of exercise in the high exercise cost state. The
manager’s option value is simply equal to the surplus the θ1-type manager receives
from the necessity of being induced into exercising at the first-best trigger P ∗1 .
3.3 Hidden Action Only Region
In the parameter range, defined by D(P ; P ∗2 )∆θ < ξ/∆q < D(P ;P ∗
1 )∆θ, the optimal
contract can be written as:
P1 = P ∗1 , (3.39)
P2 = P ∗2 , (3.40)
w1 = θ1 +ξ
∆q
(P ∗
1
P
)β
, (3.41)
w2 = θ2. (3.42)
In this region, the contract triggers equal those of the first-best outcome. The
moral hazard costs are so high that the rent needed for motivating high effort (via the
ex-ante IC constraint) is sufficiently high so that the ex-post incentive constraints do
not demand addition rents. That is, by motivating the manager to extend effort, w1
ends up being high enough so that the θ1-type manager no longer needs P2 to exceed
P ∗2 in order to dissuade him from deviating from the low-state trigger P ∗
1 . Thus, the
contract is entirely driven by the need to motivate effort, as the ex-post IC constraints
that reflect hidden information fail to bind.
We can use these contract terms to place a value on the owner’s and manager’s op-
tion values. The owner’s and manager’s option values, πo(P ) and πm(P ), respectively,
can be written as:
πo(P ) = V ∗(P ) − qHξ
∆q, (3.43)
πm(P ) = qHξ
∆q. (3.44)
The owner’s option value is equal to the first-best solution V ∗(P ) characterized in
(2.7), minus the present value of the rent paid to the manager in order to induce
18
effort. The manager’s option value is equal to the present value of this effort-inducing
compensation.
3.4 An Observationally Equivalent Setting
In the first two regions discussed, investment in the high-cost state is delayed beyond
the first-best trigger of P ∗2 . It is only in the pure hidden-action region that the first-
best trigger is achieved. In order to enhance the reader’s intuition, we provide an
observationally equivalent model that gives the same delayed investment trigger (in
the high cost state). The key point will be that increasing the costs of moral hazard
mitigates the inefficiency induced by informational asymmetry, and therefore increases
the efficiency of the investment timing decision.
Consider an alternative setting, without agency issues. However, suppose the owner
only gets a fraction δ ≤ 1 of the total project P (t) at the time of exercise. For example,
the owner may need to pay a fee proportional to the project’s value. Therefore, his
net payoff is δP (t) − θ2. Thus, the optimal trigger is simply P ∗2 /δ. The owner delays
his exercising timing, relative to the first-best (δ = 1), because the private marginal
benefit is lower than the social marginal benefit.
The exercise in each of the three regions can be replicated through an appropriately
chosen δ. In the joint hidden information/ hidden action region,
δj =P ∗
2
K2
=
(ξ
∆q∆θ
)1/βP ∗
2
P< 1. (3.45)
In the hidden information region,
δi =P ∗
2
KH
=
(1 +
qH
1 − qH
∆θ
θ2
)−1
=1 − qH
1 − qHθ1/θ2
< 1.
In the hidden action region, there is no inefficiency in exercising the option, in that
δa = 1. (3.46)
It can be seen that increasing the costs of moral hazard lead to increased investment
efficiency (although at a price of increased ex-ante effort compensation paid to the
manager). As the costs of moral hazard move us from the hidden information region
19
to the joint region, the percentage of the project retained by the firm increase, since
δj > δi in this intermediate range. As the costs of moral hazard then move us into the
hidden action region, the informational rents arising from the informational asymmetry
is smaller than the limited-liability rents induced by the hidden action. Therefore,
the owner may recommend the first-best investment timing decision to the manager
and compensate him with enough limited-liability rents. The manager will no longer
have an incentive to make costly delayed investment decision. This increase in the
“ownership stake” in the high exercise cost state represents an equivalent representation
of the timing efficiency induced by moral hazard costs.13
4 Model Implications
In this section, we analyze some of the more important implications of the optimal
contracting model. First, a clear prediction of our model is that the principal-agent
problem will introduce inertia in a firm’s investment behavior. Investment projects
are expected to be undertaken later than in the first-best setting. We thus consider
the factors that influence the expected lag in investment due to the principal-agent
problem. Second, specifically because the timing of investment differs from that of the
first-best outcome, the principal-agent problem results in a welfare loss. Obviously,
factors that make such welfare losses more pronounced will introduce an incentive for
firms and industries to alter their structures to more closely align the interests of owners
and managers.
In this section, we focus our analysis on the contract that prevails in the joint hidden
information/ hidden action region. It is in this region that the incentive problems are
the most rich and meaningful.14 Recall from Section 3 that the optimal contract in
13We are only considering increases in the costs of hidden action up to a point, equal to the rightendpoint of the hidden action region. Increases in such costs beyond this point are not supported inequilibrium, since effort will no longer be extended.
14In the hidden information region, exerting effort is worthwhile for the manager even withoutadditional payment; effectively there is really no hidden action. In the hidden action region, thereis no distortion in investment timing; hidden action and hidden information together do not lead tosocial welfare losses.
20
this joint region, D(P ;KH)∆θ ≤ ξ/∆q ≤ D(P ;P ∗2 )∆θ, is:
P1 = P ∗1 , (4.47)
P2 = K2 = P
(∆q∆θ
ξ
)1/β
, (4.48)
w1 = θ1 +
(P ∗
1
K2
)β
∆θ, (4.49)
w2 = θ2, (4.50)
where the owner’s and manager’s option values, πo(P ) and πm(P ), respectively, can be
written as:
πo(P ) = qH
(P
P ∗1
)βθ1
β − 1+ (1 − qH)
(P
K2
)β
(K2 − θ2) − qHξ
∆q, (4.51)
πm(P ) = qHξ
∆q. (4.52)
4.1 Agency Problems and Investment Lags
In the principal-agent setting, investment will occur later than it does in the standard
real options setting in which no principal-agent conflict is assumed to exist. Such in-
vestment delay due to incentive problems is consistent with the results in Bertrand
and Mullainathan (2000). In that paper, they look at the compensation and invest-
ment behavior of manufacturing firms following the passage of anti-takeover legislation.
Presumably, following such legislation, managers become more entrenched and the di-
vergence of interests between owners and managers becomes more pronounced. They
find that managers display more inertia in their investment policy following the anti-
takeover legislation. This is consistent with our model, and not as consistent with
the view that less disciplined managers behave like empire builders. They also find
that wages (including those for white collar workers) rise following the passage of the
legislation, which is also consistent with the information rents paid in our model.
In the standard real options setting, investment is triggered at the value maximizing
triggers, P ∗1 and P ∗
2 , for the low and high exercise cost outcomes, respectively. However,
in our setting, while the trigger for investment in the low exercise cost state remains at
P ∗1 , investment in the high exercise cost state may be triggered at K2, which is higher
than the first-best benchmark level P ∗2 .
21
Let T and T ∗ be the stopping times at which the option is exercised, in our model
and the first-best setting, respectively. We denote Γ = E (T − T ∗) as the expected
time lag due to the principal-agent problem. A solution for such an expectation can
be derived using Harrison (1985, Chapter 3). The expected lag is given by
Γ =
(1 − qH
α− σ2/2
)log
(K2
P ∗2
)(4.53)
=
(1 − qH
α− σ2/2
)[log
(P
θ2
)+
1
βlog
(∆q∆θ
ξ
)− log β + log (β − 1)
], (4.54)
where we assume that α > σ2/2 in order for this expectation to exist.
An important insight from Section 3 is that increases in the cost-benefit ratio of
effort leads to less distortion in investment timing. That is, as the ratio ξ/(∆q∆θ)
increases, the equilibrium trigger P2 becomes closer to the first-best trigger P ∗2 . The
intuition for this result is that as the incentive problems become more pronounced, the
owner must compensate the manager with a greater fraction of the underlying project’s
cash flows. This greater ownership stake implies that the manager’s exercise decision
will be more closely aligned with that of the owner’s. This intuition is confirmed by
the following comparative statics:
∂Γ
∂ξ= −
(1 − qH
α− σ2/2
)1
βξ< 0, (4.55)
∂Γ
∂(∆q)=
(1 − qH
α− σ2/2
)1
β∆q> 0, (4.56)
∂Γ
∂(∆θ)=
(1 − qH
α− σ2/2
)1
β∆θ> 0. (4.57)
Thus, increases in the cost of effort lead to a shorter expected investment lag, while
increases in the benefits of effort lead to a longer expected investment lag.
An increase in the volatility of the underlying project, σ2, has an ambiguous effect
on the expected time lag Γ. This can be seen from the following comparative static:
∂Γ
∂σ2= −
(1 − qH
α− σ2/2
)1
β2
[log
(∆q∆θ
ξ
)− β
β − 1
]∂β
∂σ2+
1 − qH
2 (α− σ2/2)2 log
(K2
P ∗2
),
(4.58)
where ∂β/∂σ2 < 0. An increase in σ2 raises the option value and makes waiting more
worthwhile, implying that both P ∗2 and K2 are larger, ceteris paribus. However, if the
22
marginal cost-benefit ratio for exerting effort is relatively high, in that
log
(ξ
∆q∆θ
)>β − 1
β, (4.59)
then the change of K2 relative to the change in P ∗2 is larger. Therefore, under such
conditions the expected time lag increases in σ2.
An increase in the expected growth rate of the project, α, also has an ambiguous
effect on the expected time lag Γ. This can be seen from the following comparative
static:
∂Γ
∂α= − 1 − qH
(α− σ2/2)2
log
(K2
P ∗2
)− 1
β
(log
(∆q∆θ
ξ
)− β
β − 1
)α− σ2/2√
(α− σ2/2)2 + 2rσ2
.
(4.60)
However, if (4.59) holds, then expected time lag decreases with drift α.
4.2 Welfare Loss and Option Values
Although the owner chooses the value-maximizing contract to provide an incentive for
the manager to extend effort, the agency problem ultimately still proves costly. In an
owner-managed firm, the manager will extend effort and will exercise the option at the
first-best stopping times. However, in firms with delegated management, there will be
a welfare loss due to the firm’s suboptimal exercise strategy.
By a welfare loss, we are referring to the difference between the values of the first-
best option value, V ∗(P ) in (2.9), and the sum of the owner and manager options,
πo(P ) and πm(P ) in (4.51) and (4.52). Thus, define the welfare loss due to agency
issues as L, where L = V ∗(P ) − (πo(P ) + πm(P )). Simplifying, we have:
L = (1 − qH)
[(P
P ∗2
)β
(P ∗2 − θ2) −
(P
K2
)β
(K2 − θ2)
]. (4.61)
This welfare loss is likely to have economic ramifications on the structure of firms.
For firms in industries with potentially large welfare losses due to agency, there will
be powerful forces that will push them to be privately held, or to be organized in a
manner that provides the closest alignment between owners and managers.
The size of the welfare loss is driven by the distance of the equilibrium trigger
K2 from the first-best trigger P ∗2 . As previously discussed, the firm’s exercise timing
23
becomes less distorted as the net cost-benefit ratio of effort increases. That is, as the
ratio ξ/( ∆q∆θ) increases, the equilibrium triggerK2 gets closer to the first best trigger
P ∗2 , and thus:
∂L
∂ξ< 0, (4.62)
∂L
∂∆q> 0, (4.63)
∂L
∂∆θ> 0. (4.64)
In terms of the owner’s option value, the incentive problem represents a trade-off
between timing efficiency and the surplus that must be paid to the manager to extend
effort. One can obtain better intuition on the forces at work in the agency problem
through the following decomposition. In the first-best solution, the owner pays the cost
of effort ξ and obtains the first-best option value V ∗(P ). In the agency equilibrium,
the owner delegates the cost of effort to the manager, but then holds the sub-optimal
option value πo(P ). The loss in the owner’s option value due to the incentive problem
is therefore given by:
∆πo(P ) ≡ V ∗(P ) − ξ − πo(P ) = L+ V m, (4.65)
where L is the total welfare loss given in (4.61), and V m is the ex-ante expected surplus
paid to the manager to exert effort, and is given by:
V m = πm(P ) − ξ = qH
(P
P ∗1
)β
(w1 − θ1) − ξ = qHξ
∆q− ξ =
qL
∆qξ. (4.66)
The decomposition of the welfare cost to the owner from (4.65) into the sum of
the timing component (L) and the compensation component (V m), is very useful for
providing intuition. The impact of a higher effort cost ξ represents a trade-off in
terms of the timing and compensation components. As shown in (4.62), a higher effort
cost results in more efficient investment timing. This must be traded-off against the
increased surplus that must be paid to provide appropriate incentives to the manager,
as seen in (4.66). Therefore, the total effect on the loss of owner’s option value due
24
to an increase in ξ depends on whether the “timing effect” or “compensation effect” is
larger, in that
∂
∂ξ∆πo(P ) = − (1 − qH) (β − 1)
(P
K2
)β (1 − P ∗
2
K2
)K2
βξ+qL
∆q(4.67)
=β − 1
β
1
∆q∆θ[− (1 − qH) (K2 − P ∗
2 ) + qL (P ∗2 − P ∗
1 )] . (4.68)
If the investment trigger K2 is significantly larger than P ∗2 , in that
(1 − qH) (K2 − P ∗2 ) > qL (P ∗
2 − P ∗1 ) , (4.69)
then an increase in ξ leads to a smaller loss in the owner’s option value, as the gain in
timing efficiency overcomes the loss due to the manager’s incentive surplus.15
5 Impatient Managers and Earlier-than-Optimal In-
vestment
In the standard model, both owners and managers value payoffs identically. However,
it may be the case that owners and managers have different discount rates. A manager
very well may have a higher discount rate for the project’s cash flows than the owner.
The manager’s higher discount rate will be result in greater impatience; the manager
will be more willing than the owner to take a reduced future payoff in exchange for
receiving it earlier. The most interesting result of this generalization is that the
manager may now invest earlier than the first-best optimum. Thus, while the standard
model results in a form of under-investment, in the generalized model we can have both
over and under-investment.
Consider the following potential motivations for a manager possessing a higher
discount rate than the owner. For example, even if shocks to the underlying project’s
returns are solely idiosyncratic and uncorrelated with the market, an undiversified
15Note that the above condition is non-empty. This can be seen as follows. Condition 4.69 isequivalent to
K2 >1
1 − qH[(1 − qH)P ∗
2 + qL(P ∗2 − P ∗
1 )] = KH − ∆q
1 − qH(P ∗
2 − P ∗1 ) .
The joint moral hazard/hidden information region is characterized by P ∗2 ≤ K2 ≤ KH . Therefore,
the above condition is met for some K2.
25
manager may discount project cash flows at a rate greater than the riskless rate. As
another example, the manager may use a higher discount rate if his effective horizon
is shorter. More specifically, suppose the manager faces an exogenous termination
driven by a Poisson process with intensity λ, where termination may occur at any
time prior to the realization of P1. This termination could be due to the manager’s
death, finding another job, or forced termination, but is not explicitly modeled.16 The
addition of stochastic termination transforms the manager’s option to one in which his
discount rate r is elevated to r + λ, a higher discount rate, to reflect the stochastic
termination.17
In this section, we consider a setting where the manager and the owner have different
discount rates. We suppose that the manager is more impatient than the owner.
Mathematically, this implies separate discount functions for the owner and manager,
Do(P ; P ) and Dm(P ; P ), where P ≤ P :
Do(P ; P ) =
(P
P
)β
, (5.70)
Dm(P ; P ) =
(P
P
)γ
. (5.71)
Since ∂β/∂r > 0, the manager’s higher discount rate is embodied by the condition
γ > β.
The problem is basically the same as that of Section 2.3, with the exception that
the constraints all use the higher discount factor γ rather than β. In addition, much
of the solution is similar. For example, Lemmas 1 and 2 apply as before, using the
same proof. In addition, Lemma 3 also holds, and is demonstrated in the appendix.
Thus, the optimal contracting problem in the generalized setting can be written as:
maxw1, w2, P1, P2
(P
P1
)β
(P1 − w1) +1 − qH
qH
(P
P2
)β
(P2 − w2) (5.72)
16We assume that the owner can costlessly replace the manager in the event of separation.17See Yaari (1965), Merton (1971) and Richard (1975) for this result.
26
subject to (P
P1
)γ
(w1 − θ1) ≥(P
P2
)γ
(θ2 − θ1) (5.73)
0 ≥(P
P1
)γ
(w1 − θ2) (5.74)(P
P1
)γ
(w1 − θ1) ≥ ξ
∆q. (5.75)
Just as in Section 3, there are three contracting regions: a hidden information
region, a joint hidden information-hidden action region, and a hidden action region.
In this section, we focus on the joint hidden information/hidden action region, as it
provides the richest setting in which to analyze the model.18
The joint hidden information-hidden action region is defined by the following region:(P
KH
)γ
∆θ < ξ∆q
<(
PP ∗
2
)γ
∆θ, where KH > P ∗2 is defined in (B.25). In this region the
optimal contract can be written as:
P1 = K1, (5.76)
P2 = K2 = P
(∆q∆θ
ξ
)1/γ
, (5.77)
w1 = θ1 +
(P1
K2
)γ
∆θ < θ2, (5.78)
w2 = θ2, (5.79)
where K1 is the root of H(x) = 0, where:
H(x) =β
β − 1
[θ1 +
(1 − γ
β
)( xP
)γ ξ
∆q
]− x. (5.80)
Importantly, unlike the basic model, we now have the possibility of investment
occurring before the first-best trigger, in that P1 < P ∗1 . To see this, note that H(0) =
P ∗1 and
H(P ∗1 ) =
β
β − 1
(1 − γ
β
)(P ∗
1
P
)γξ
∆q< 0. (5.81)
The derivative of H( · ) is
H ′(x) =β
β − 1γ
(1 − γ
β
)( xP
)γ−1 1
P
ξ
∆q− 1 < 0. (5.82)
18The derivations for the optimal contracts in the other regions are shown in the appendix.
27
Therefore, there exists a unique solution P1 = K1 < P ∗1 .
As in the standard model, the trigger in the high exercise state is greater than
the first-best trigger, P ∗2 . This is based on the fact that K2 > P ∗
2 in the region
Dm(P ; KH)∆θ < ξ/∆q < Dm(P ;P ∗2 )∆θ. However, for γ > β, the trigger is closer to
the first-best trigger than for the standard case in which γ = β. This is true, since for
γ > β,
K2 = P
(∆q∆θ
ξ
)1/γ
< P
(∆q∆θ
ξ
)1/β
= K2. (5.83)
Thus, we have the result that in a model in which the manager is more impatient
than the owner, equilibrium investment occurs sooner than it does in the standard
model, with investment actually occurring prior to the first-best trigger in the low
exercise cost state. The greater impatience on the part of the manager implies that it
is in the owner’s interest to offer a contract that motivates earlier exercise. This results
in both costs and benefits to the owner. By motivating earlier investment in the high-
cost state, investment timing moves closer to first-best. Since the manager receives no
surplus in this state, the owner is the sole beneficiary of this timing efficiency. However,
earlier investment in the low-cost state implies investment that occurs sooner than the
first-best outcome. In this state, the owner is worse off for two reasons: investment
occurs too early, and the wage paid to the manager in this state must be higher (than
in the standard model) in order to motivate early investment. The net effect of these
costs and benefits is ambiguous and is driven by the relative parameter values.
6 Concluding Remarks
This paper extends the real options framework to account for the agency issues that are
prevalent in many real-world applications. When investment decisions are delegated
to managers, contracts must be designed that provide an incentive for the manager to
both extend effort and to exercise optimally. This article provides a model of optimal
contracting in a continuous-time principal-agent setting in which there is both moral
hazard and adverse selection. The implied investment behavior differs significantly from
that of the first-best no-agency solution. In particular, there will be greater inertia in
investment, as the model leads to the manager having an even greater “option to wait”
28
than the owner. The interplay between the twin forces of hidden information and
hidden action leads to markedly different investment outcomes than when only one of
the two forces is at work. Allowing the manager to have an effort choice that affects the
likelihood of getting a high quality project mitigates the investment inefficiency due to
informational asymmetry. When the model is generalized to include differing degrees
of impatience between owners and managers, we find that investment may occur either
earlier or later than optimal.
Some extensions of the model would prove interesting. First, the model could allow
for repeated investment decisions. This richer setting would permit owners to update
their beliefs over time, and for managers to establish reputations. Second, the model
could also be generalized to include competition in both the labor and product markets.
As shown by Grenadier (2002), the forces of competition greatly alter the investment
behavior implied by real options models.
29
Appendices
A Solution to the Optimal Contracting Problem
This appendix provides a derivation of the optimal contracts detailed in Section 3.
We proceed by first proving Lemma 3. Then, we present the simplified optimization
problem and derive the optimal contract in the three critical regions.
The owner’s objective is to maximize the value of his option, as expressed in (2.13).
As discussed in Section 2.3, this optimization is subject to the ex-ante and ex-post
incentive constraints (2.14), (2.15), (2.16), (2.17) and (2.18). Lemmas 1 and 2 allow
us to simplify the problem as follows:
maxw1, w2, P1, P2
(P
P1
)β
(P1 − w1) +1 − qH
qH
(P
P2
)β
(P2 − w2), (A.1)
subject to (P
P1
)β
(w1 − θ1) ≥(P
P2
)β
(w2 − θ1) (A.2)
(P
P2
)β
(w2 − θ2) ≥(P
P1
)β
(w1 − θ2) (A.3)
(P
P1
)β
(w1 − θ1) −(P
P2
)β
(w2 − θ2) ≥ ξ
∆q(A.4)
w2 ≥ θ2. (A.5)
Using the method of Kuhn-Tucker, we form the Lagrangian as follows:
L =
(P
P1
)β
(P1 − w1) +1 − qH
qH
(P
P2
)β
(P2 − w2)
+ λ1
[(P
P1
)β
(w1 − θ1) −(P
P2
)β
(w2 − θ1)
]
+ λ2
[(P
P2
)β
(w2 − θ2) −(P
P1
)β
(w1 − θ2)
]
+ λ3
[(P
P1
)β
(w1 − θ1) −(P
P2
)β
(w2 − θ2) − ξ
∆q
]+ λ4 (w2 − θ2) , (A.6)
with corresponding complementary slackness conditions for the four constraints. The
first-order condition with respect to w1 implies:
λ1 − λ2 + λ3 = 1. (A.7)
30
The first-order condition with respect to w2 implies:(−λ1 + λ2 − λ3 − 1 − qH
qH
)(P
P2
)β
+ λ4 = 0. (A.8)
Using (A.7) to simplify (A.8),
λ4 =
(P
P2
)β1
qH> 0. (A.9)
Therefore, the complementary slackness condition λ4(w2−θ2) = 0 implies that w2 = θ2.
This proves Lemma 3.
Using Lemmas 1, 2 and 3, we can now write the simplified problem as:
maxw1, P1, P2
(P
P1
)β
(P1 − w1) +(1 − qH)
qH
(P
P2
)β
(P2 − θ2) (A.10)
subject to (P
P1
)β
(w1 − θ1) ≥(P
P2
)β
(θ2 − θ1) (A.11)
0 ≥(P
P1
)β
(w1 − θ2) (A.12)
(P
P1
)β
(w1 − θ1) ≥ ξ
∆q. (A.13)
The first-order conditions with respect to P1 and P2 imply:
P1 =β
β − 1(θ1 − λ2∆θ) , (A.14)
P2 =β
β − 1
(θ2 +
qH
1 − qH
λ1 ∆θ
). (A.15)
We conjecture that the ex-post IC constraint for the type-θ2 manager, (A.12), does
not bind, in that λ2 = 0, for all three regions that we analyze. We will verify this
conjecture for each region. By combining the conjecture that λ2 = 0 with Lemma 3,
we have that P1 = P ∗1 and w2 = θ2 in each of the relevant contract regions.
With λ2 = 0, then (A.7) may be written as λ1 + λ3 = 1. Therefore, it must be the
case that at least one of (A.11) and (A.13) binds. This is consistent with our intuition.
Otherwise, lowering w1 increases the expected surplus to the owner, without violating
any other constraints.
31
A.1 The Joint Hidden Information-Hidden Action Region:(P
KH
)β
∆θ ≤ ξ∆q ≤
(PP ∗
2
)β
∆θ
We derive the optimal contract in this region by conjecturing that both (A.11) and
(A.13) bind. Solving these two equality constraints gives us:
P2 = K2 = P
(∆q∆θ
ξ
)1/β
, (A.16)
w1 = θ1 +
(P ∗
1
K2
)β
∆θ.
The solution for P2 implies that λ1 can be written as:
λ1 =β − 1
β(K2 − P ∗
2 )1 − qH
qH∆θ. (A.17)
The only possible regions under which both constraints may bind19 is characterized by(P
KH
)β
∆θ <ξ
∆q<
(P
P ∗2
)β
∆θ. (A.18)
The above argument implies that the only possible region under which both constraints
hold must lie within (A.18). But it does not necessarily imply that all the region will
require that both constraints bind. However, we now show that both (A.11) and
(A.13) bind throughout this entire region. The region characterized by (A.18) can be
equivalently expressed as P ∗2 < K2 < KH . Because (A.17) implies that λ1 is monotonic
increasing in K2, therefore, the Lagrangian multiplier is characterized by 0 < λ1 < 1,
in this region. Since λ3 = 1 − λ1, we also have 0 < λ3 < 1. Therefore, by the
complementary slackness conditions, both (A.11) and (A.13) bind in this joint region,
confirming the result that (A.18 is the whole region, with both constraints binding.
A.2 The Hidden Information Region: ξ∆q <
(P
KH
)β
∆θ
Suppose that the ex-ante IC constraint (A.13) does not bind. Since (A.11) must hold
as an equality, and since λ1 = 1, we therefore have
P2 = KH =β
β − 1
(θ2 +
qH
1 − qH
∆θ
), (A.19)
19If(
PP ∗
1
)β
∆θ > ξ∆q >
(PP ∗
2
)β
∆θ, then only the third constraint binds. If(
PKH
)β
∆θ > ξ∆q , then
only the first constraint binds. If ξ∆q >
(PP ∗
1
)β
∆θ, then supporting high effort is no longer in theowner’s interest.
32
and
w1 = θ1 +
(P ∗
1
KH
)β
∆θ < θ2. (A.20)
The inequality in (A.20) confirms that (A.12) does not bind, as conjectured. Finally,
in order to be consistent with the assumption that (A.13) doesn’t bind, we require that
ξ
∆q<
(P
KH
)β
∆θ. (A.21)
This is indeed confirmed by the definition of this contract region.
A.3 The Hidden Action Region:(
PP ∗
2
)β
∆θ < ξ∆q <
(PP ∗
1
)β
∆θ
Suppose that (A.11) does not bind and (A.13) binds, then λ1 = 0 by complementary
slackness, and λ3 = 1. Therefore,
P2 = P ∗2 , (A.22)
w1 = θ1 +
(P ∗
1
K2
)β
∆θ.
We need to verify that (A.11) and (A.12) do not bind. The constraint (A.12) is
non-binding if and only if P ∗1 < K2. The constraint (A.11) is non-binding if and only
if K2 < P ∗2 . Thus, together this implies that P ∗
1 < K2 < P ∗2 , which is identical to the
condition(
PP ∗
2
)β
∆θ < ξ∆q
<(
PP ∗
1
)β
∆θ that defines this region.
If the parameters do not fall in any of the three regions, namely, ξ∆q
>(
PP ∗
1
)β
∆θ,
then it can be shown that the owner will not choose to motivate the manager to exert
effort. The cost of effort is so high as to overwhelm any potential benefits of motivating
effort. A proof of this result is available from the authors upon request.
B Solution to the Generalized Optimal Contracting
Problem in Section 5
This appendix provides a derivation of the optimal contract detailed in Section 5. Since
Lemma 1 and Lemma 2 apply as in the standard model, the owner solves the following
optimization problem:
maxw1, w2, P1, P2
(P
P1
)β
(P1 − w1) +1 − qH
qH
(P
P2
)β
(P2 − w2) (B.1)
33
subject to (P
P1
)γ
(w1 − θ1) ≥(P
P2
)γ
(w2 − θ1) (B.2)(P
P2
)γ
(w2 − θ2) ≥(P
P1
)γ
(w1 − θ2) (B.3)(P
P1
)γ
(w1 − θ1) −(P
P2
)γ
(w2 − θ2) ≥ ξ
∆q(B.4)
w2 ≥ θ2. (B.5)
Using the method of Kuhn-Tucker, we form the Lagrangian as follows:
L =
(P
P1
)β
(P1 − w1) +1 − qH
qH
(P
P2
)β
(P2 − w2)
+ λ1
[(P
P1
)γ
(w1 − θ1) −(P
P2
)γ
(w2 − θ1)
]
+ λ2
[(P
P2
)γ
(w2 − θ2) −(P
P1
)γ
(w1 − θ2)
]
+ λ3
[(P
P1
)γ
(w1 − θ1) −(P
P2
)γ
(w2 − θ2) − ξ
∆q
]+ λ4 (w2 − θ2) . (B.6)
with corresponding complementary slackness conditions for the four constraints.The
first-order condition with respect to w1 implies:
(λ1 − λ2 + λ3)
(P
P1
)γ
=
(P
P1
)β
. (B.7)
The first-order condition with respect to w2 implies:
(−λ1 + λ2 − λ3)
(P
P2
)γ
− 1 − qH
qH
(P
P2
)β
+ λ4 = 0. (B.8)
Using (B.7) to simplify (B.8),
λ4 =
(P
P2
)β[(
P2
P1
)β−γ
+1 − qH
qH
]> 0. (B.9)
The complementary slackness condition implies that w2 = θ2. This proves Lemma 3
for the generalized model.
Therefore, we may simplify the optimization problem as follows:
maxw1, P1, P2
(P
P1
)β
(P1 − w1) +1 − qH
qH
(P
P2
)β
(P2 − θ2) (B.10)
34
subject to (P
P1
)γ
(w1 − θ1) ≥(P
P2
)γ
(θ2 − θ1) (B.11)
0 ≥(P
P1
)γ
(w1 − θ2) (B.12)(P
P1
)γ
(w1 − θ1) ≥ ξ
∆q. (B.13)
The first-order condition with respect to P1 is given by
P1 =β
β − 1
[w1 − γ
β[(λ1 + λ3) (w1 − θ1) − λ2 (w1 − θ2)]
(P
P1
)γ−β]. (B.14)
The first-order condition with respect to P2 is given by
P2 =β
β − 1
[θ2 + λ1
qH
1 − qH
γ
β
(P
P2
)γ−β
∆θ
]. (B.15)
Also note that
λ1 + λ3 = λ2 +
(P
P1
)β−γ
. (B.16)
We conjecture that the ex-post IC constraint for the type-θ2 manager, (B.12), does not
bind, in that λ2 = 0, for all three regions that we analyze. We will verify this conjecture
for each region. With λ2 = 0, then (B.16) may be written as λ1 + λ3 =(
PP1
)β−γ
> 1.
Therefore, it must be the case that at least one of (B.11) and (B.13) binds. This is
consistent with our intuition. Otherwise, lowering w1 increases the expected surplus
to the manager, without violating any other constraints.
B.1 The Hidden Information Region: ξ∆q <
(P
KH
)γ
∆θ
Suppose that the third constraint (B.13) does not bind and thus λ3 = 0. Then,
λ1 =
(P
P1
)β−γ
> 1. (B.17)
A binding ex-ante IC constraint (B.11) implies that the wage payment is
w1 = θ1 +
(P1
P2
)γ
∆θ. (B.18)
35
Using the first-order conditions (B.14) and (B.15), the optimal trigger levels are
P1 =β
β − 1
[θ1 +
(1 − γ
β
)(P1
P2
)γ
∆θ
](B.19)
P2 =β
β − 1
[θ2 +
qH
1 − qH
γ
β
(P1
P2
)γ−β
∆θ
]. (B.20)
Note that with γ > β, we immediately have P1 < P ∗1 and P2 > P ∗
2 . Therefore, w1 < θ2,
as conjectured, confirming that (B.12) does not bind.
Note that the system of equations (B.19) and (B.20) can be solved as follows. The
ratio x∗ = P1/P2 solves
G(x) = 0, (B.21)
where
G(x) = x
[θ2 +
γ
β
(qH
1 − qH
)xγ−β∆θ
]−[θ1 +
(1 − γ
β
)xγ∆θ
]= 0. (B.22)
First, note that G(0) = −θ1 < 0 and G(1) = γ/(β(1 − qH)) > 0. Second,
G′(x) = θ2 +γ + γ(γ − β)
β
qH
1 − qH
xγ−β∆θ + γxγ−1
(γ
β− 1
)∆θ > 0, (B.23)
for γ > β. Therefore, there exist a unique x∗ ∈ (0, 1) solving (B.21).
Therefore, for the region defined by ξ∆q
<(
PKH
)γ
∆θ, the optimal contract can be
written as:
P1 =β
β − 1
[θ1 +
(1 − γ
β
)(x∗)γ ∆θ
], (B.24)
P2 = KH =β
β − 1
[θ2 +
qH
1 − qH
γ
β(x∗)γ−β∆θ
], (B.25)
w1 = θ1 +
(P1
P2
)γ
∆θ, (B.26)
w2 = θ2. (B.27)
Finally, since in this region(
PKH
)γ
∆θ > ξ∆q
, constraint (B.13) is indeed not binding.
B.2 The Hidden Action Region:(
PP ∗
2
)γ
∆θ < ξ∆q <
(PK1
)γ
∆θ
Suppose that (B.13) binds, while (B.11) does not. Thus, λ1 = 0 and
λ3 =
(P
P1
)β−γ
> 1. (B.28)
36
Since λ1 = 0, (B.14) implies that
P2 = P ∗2 . (B.29)
Substituting λ3 =(
PP1
)β−γ
into (B.15) results in:
P1 = K1 =β
β − 1
[θ1 +
(1 − γ
β
)(P1
P
)γξ
∆q
]=
β
β − 1
[θ1 +
(1 − γ
β
)(P1
K2
)γ
∆θ
].
(B.30)
In Section 5 we prove that a unique K1 exists, where K1 ∈ (0, P ∗1 ).
A binding (B.13) implies that the wage payment is
w1 = θ1 +
(K1
P
)γξ
∆q= θ1 +
(K1
K2
)γ
∆θ, (B.31)
where K2 = P(
∆q∆θξ
)1/γ
. To ensure are conjecture that (B.12) doesn’t bind, (B.31)
implies that we need K1 < K2. This inequality can be written as ξ∆q
<(
P
K1
)γ
∆θ,
which is thus assured to hold in this region.
In order to be consistent with the fact that (B.11) does not bind, we need(
PP ∗
2
)γ
∆θ >ξ
∆q, which holds in this region.
Therefore, for the region defined by(
PP ∗
2
)γ
∆θ < ξ∆q
<(
PK1
)γ
∆θ, the optimal
contract can be written as:
P1 = K1, (B.32)
P2 = P ∗2 , (B.33)
w1 = θ1 +
(K1
K2
)γ
∆θ, (B.34)
w2 = θ2. (B.35)
B.3 The Joint Hidden Information/Hidden Action Region:(P
KH
)γ
∆θ ≤ ξ∆q ≤
(PP ∗
2
)γ
∆θ
We derive the optimal contract in this region by conjecturing that both (B.11) and
(B.13) bind. Solving these two equality constraints gives us:
P2 = K2 = P
(∆q∆θ
ξ
)1/γ
, (B.36)
w1 = θ1 +
(P1
K2
)γ
∆θ. (B.37)
37
Plugging the above two equations into the first-order condition (B.14) gives
P1 = K1 =β
β − 1
[θ1 +
(1 − γ
β
)(P1
K2
)γ
∆θ
], (B.38)
the same solution for P1 as in the hidden action region. In Section 5 we prove that a
unique K1 exists, where K1 ∈ (0, P ∗1 ). In addition, w1 in (B.37) is the same solution
as in the hidden action region. Thus, as before, we have verified that (B.12) does not
bind in this region.
We know that the only possible regions in which both (B.11) and (B.13) bind is(P
KH
)γ
∆θ < ξ∆q
<(
PP ∗
2
)γ
∆θ, since we have already shown that in the other regions
only one of these constraints binds.20 Equivalently stated in terms of K2, this region
is characterized by P ∗2 < K2 < KH . We now verify that the above solutions are indeed
optimal for this entire region.
In order to confirm that both the first and third constraints bind, we require that
λ1, λ3 �= 0. Since λ2 = 0 in this entire region, λ1 + λ3 =(
PP1
)β−γ
. Thus, we require
that
0 < λ1 <
(P
K1
)β−γ
. (B.39)
The first-order condition with respect to P2 implies that
λ1 =
[β
β − 1
qH
1 − qH
γ
β
(P
K2
)γ−β
∆θ
]−1 (K2 − P ∗
2
)(B.40)
Since K2 > P ∗2 , we have confirmed that λ1 > 0. We now prove that λ1 <
(PK1
)β−γ
.
We now consider λ1 as a function of K2, and rewrite (B.40) as:
λ1(K2) =
(P
K1(K2)
)β−γ (K1(K2)
K2
)β−γ [qH
1 − qH
γ
β(P ∗
2 − P ∗1 )
]−1 (K2 − P ∗
2
).
(B.41)
Note that from (B.38), K1 is a function of K2; we make this functional dependence
explicit in the above equation.
20Note that in the region ξ∆q >
(PK1
)γ
∆θ, it can be shown that effort cannot be induced. Thisresult is available from the authors by request.
38
We need to show that λ1(K2) <(
P
K1(K2)
)β−γ
over the region K2 ∈ (P ∗2 , KH). This
is equivalent to showing that
K2 < P ∗2 +
(K1(K2)
K2
)γ−βqH
1 − qH
γ
β(P ∗
2 − P ∗1 ). (B.42)
Let
L(x) = P ∗2 +
(K1(x)
x
)γ−βqH
1 − qH
γ
β(P ∗
2 − P ∗1 ) − x. (B.43)
We thus need to show that L(x) > 0 for x ∈ (P ∗2 , KH).
Using implicit differentiation in (B.38), we can write:
dK1(x)
dx=K1(x)
x
γ(K1(x) − P ∗1 )
γ(K1(x) − P ∗1 ) − K1(x)
> 0, (B.44)
because K1(x) < P ∗1 in this region. Therefore,
dL(x)
dx= (γ − β)
qH
1 − qH
γ
β(P ∗
2 − P ∗1 )
(K1(x)
x
)γ−β−1(xdK1(x)
dx− K1(x)
x2
)− 1. (B.45)
From (B.44),
xdK1(x)
dx− K1(x) =
[K1(x)
]2γ(K1(x) − P ∗
1 ) − K1(x)< 0, (B.46)
because K1(x) < P ∗1 in this region. Therefore, dL(x)
dx< 0 for x ∈ (P ∗
2 , KH). Since
L(KH) = 0, we thus have L(x) > 0 for x ∈ (P ∗2 , KH). This confirms that λ1, λ3 > 0
in this entire region, and therefore both (B.11) and (B.13) bind.
Therefore, for the region defined by(
P
KH
)γ
∆θ ≤ ξ∆q
≤(
PP ∗
2
)γ
∆θ, the optimal
contract can be written as:
P1 = K1, (B.47)
P2 = K2, (B.48)
w1 = θ1 +
(K1
K2
)γ
∆θ, (B.49)
w2 = θ2. (B.50)
39
References
[1] Bernardo, A. E., H. Cai, and J. Luo, 2001, “Capital Budgeting and Compen-
sation with Asymmetric Information and Moral Hazard,” Journal of Financial
Economics 61, 311-344.
[2] Bertrand, M., and S. Mullainathan, 2000, “Enjoying the Quiet Life: Manager-
ial Behavior Following Anti-Takeover Legislation,” Journal of Political Economy,
forthcoming.
[3] Brennan, M.J., and E. Schwartz, 1985, “Evaluating Natural Resource Invest-
ments,” Journal of Business 58, 135-157.
[4] Dixit, A., 1989, “Entry and Exit Decisions Under Uncertainty,” Journal of Polit-
ical Economy 97, 620-638.
[5] Dixit, A. K., and R. S. Pindyck, 1994, “Investment Under Uncertainty,” Princeton
University Press, Princeton, New Jersey.
[6] Grenadier, S. R., 1995, “The Persistence of Real Estate Cycles,” Journal of Real
Estate Finance and Economics 10, 95-119.
[7] Grenadier, S. R., 2002, “Option Exercise Games: An Application to the Equilib-
rium Investment Strategies of Firms,” Review of Financial Studies 15, 691-721.
[8] Hendricks, K., and D. Kovenock, 1989, “Asymmetric Information, Information
Externalities, and Efficiency: The Case of Oil Exploration,” RAND Journal of
Economics 20, 164-182.
[9] Hendricks, K., R. H. Porter, and B. Boudreau, 1987, “Information and Returns
in OCS Auctions: 1954-1969,” Journal of Industrial Economics 35, 517-542.
[10] Holmstrom, B. R., 1979, “Moral Hazard and Observability,” Bell Journal of Eco-
nomics 10, 74-91.
[11] Innes, R. D., 1990, “Limited Liabilities and Incentive Contracting with Ex-ante
Action Choices,” Journal of Economic Theory 52, 45-67.
40
[12] Institute of Real Estate Management Foundation and Arthur Andersen, 1991,
“Managing the Future: Real Estate in the 1990s.”
[13] Maeland, J., 2002, “Asymmetric Information About a Constant Investment Cost,”
working paper, Norwegian School of Economics.
[14] Majd, S., and R. Pindyck, 1987, “Time to Build, Option Value, and Investment
Decisions,” Journal of Financial Economics 18, 7-28.
[15] McDonald, R., and D. Siegel, 1986, “The Value of Waiting to Invest,” Quarterly
Journal of Economics 101, 707-727.
[16] Merton, R. C., 1973, “Theory of Rational Option Pricing,” Bell Journal of Eco-
nomics and Management Science 4, 141-183.
[17] Merton, R. C., 1971, “Optimal Consumption and Portfolio Rules in a Continuous-
Time Model,” Journal of Economic Theory 3, 373-413.
[18] Mirrlees, J., 1976, “The Optimal Structure of Incentives and Authority Within an
Organization,” Bell Journal of Economics 7, 105-131.
[19] Paddock, J. L., D. R. Siegel, and J. L. Smith, 1988, “Option Valuation of Claims
on Real Assets: The Case of Offshore Petroleum Leases,” Quarterly Journal of
Economics 103, 479-508.
[20] Richard, S., 1975, “Optimal Consumption, Portfolio, and Life Insurance Rules for
an Uncertain Lived Individual in a Continuous-Time Model,” Journal of Financial
Economics 2, 187-204.
[21] Titman, S., 1985, “Urban Land Prices Under Uncertainty,” American Economic
Review 75, 505-514.
[22] Triantis, A., and J. Hodder, 1990, “Valuing Flexibility as a Complex Option,”
Journal of Finance 45, 549-565.
[23] Trigeorgis, L., 1996, “Real Options,” MIT Press, Cambridge, MA.
41
[24] Williams, J. T., 1991, “Real Estate Development as an Option,” Journal of Real
Estate Finance and Economics 4, 191-208.
[25] Williams, J. T., 2001, “Agency, Ownership, and Returns on Real Assets,” working
paper, Professors Capital.
[26] Yaari M. E., 1965, “Uncertain Lifetime, Life Insurance, and the Theory of the
Consumer,” Review of Economic Studies 32, 137-150.
42