You are on page 1of 47

Ethics and Agency Theory:

Incorporating a Standard for Effort and an Ethically Sensitive Agent





Douglas E. Stevens
Syracuse University
M. J. Whitman School of Management
Syracuse, NY 13244-2130
Phone: 315-443-3587
Fax: 315-443-5457
E-mail: desteven@syr.edu



Alex Thevaranjan
Syracuse University
M. J. Whitman School of Management
Syracuse, NY 13244-2130
Phone: 315-443-3355
Fax: 315-443-5457
E-mail: athevara@syr.edu



November 25, 2003



This research was funded through a grant from the Syracuse University School of Management
Research Fund. The authors would like to acknowledge the helpful comments of Anwer S.
Ahmed, Amiya K. Basu, John H. Evans III, Gerald A. Feltham, Paul E. Fischer, David G. Harris,
Steven J. Huddart, Gerald J. Lobo, Eric W. Noreen, Brian P. Shapiro, Mary Stanford and
workshop participants at the Sixth Annual Professional and Ethics Symposium, the University of
British Columbia, Syracuse University, and the University of Washington.


Ethics and Agency Theory:
Incorporating a Standard for Effort and an Ethically Sensitive Agent

We study the implications of introducing ethics into the traditional principal-
agent model. In our model, the principal specifies a standard for effort at the time
of contracting and the agent suffers a utility loss if he chooses not to provide the
standard after agreeing to the contract. The magnitude of the loss depends upon
the agents ethical sensitivity. We demonstrate the emergence of an optimal flat
salary contract. We then examine the interplay between ethical sensitivity and
firm productivity in determining the optimal salary contact, and contrast it with
the traditional incentive solution. Our results are intuitive and help explain a
variety of contracting behavior that is inconsistent with traditional agency
predictions. (JEL: A31, D82, M14)

Researchers in accounting, finance, and economics have frequently used agency theory to
study issues in organizational control (See reviews by Stanley Baiman 1982, John W. Pratt and
Richard J. Zeckhauser 1985, Kathleen M. Eisenhardt 1989, Baiman 1990, and Barry M. Mitnick
1992). Traditional agency models, however, assume that individuals are opportunistic and
motivated solely by economic self-interest. That is, individuals make choices that maximize their
own economic utility independent of the utility of others or abstract values such as honesty, duty
or fairness. Based on this view of human behavior, agency models prescribe complex incentive
schemes and costly monitoring to control opportunistic behavior within the organization.
Employment contracts found in practice, however, bear little resemblance to those
predicted by the theory (Joseph E. Stiglitz 1991). Such contracts are simple and incomplete, and
1

the level of monitoring is far less than that required by the theory. This state of affairs has led
researchers to question why agents dont shirk more in the absence of optimal contracts. Herbert
A. Simon (1991) asserts that employees exert effort because they identify with the goals of the
firm, and attributes this to a genetically based propensity for docility. Timothy Besley and
Maitreesh Ghatak (2003) describe workers as mission-oriented agents who work hard when
the goals of the firm match their own. Other researchers, however, see the employment contracts
found in practice as evidence of ethical values at work (Kenneth J. Arrow 1985). Kenneth
Koford and Mark Penno (1992) assert that most people have attitudes toward telling the truth
and providing fair amounts of effort, and agency models neglect a significant element of reality
by failing to incorporate such attitudes.
Researchers have already begun to study the potential effect of ethics on common
economic relationships. Eric W. Noreen (1988) uses an economic framework to show how
ethical behavior makes the formation of markets and organizations possible. Thomas H. Noe and
Michael J. Rebello (1994) demonstrate more formally how ethical norms increase the
profitability of investment opportunities. Koford and Penno (1992) model ethical behavior in
various ways to show how the results could better reflect behavior within the organization.
1

Recent experimental studies also support the potential role of ethics in organizational control.
Human subjects have been found to give up some earnings in order to honestly report their
production potential (J. Harry Evans III et al. 2001), reduce budgetary slack (Douglas E. Stevens
2002), or provide higher effort (Jeffrey W. Schatzberg and Stevens 2003).
Despite the potential for enhancing descriptive validity, however, agency theorists have
not explicitly modeled the effect of ethics on the contracting behavior of principals and agents.
Mitnick (1992) conjectures that the mathematical formalism that has developed in the principal-
2

agent literature in accounting, finance, and economics has closed it to influences from the ethics
literature.
2
This is unfortunate, since the notion of potentially offsetting preferences for ethics
and earnings has existed from the inception of economic theory (Adam Smith 1759) and has
received particular attention in other areas of the literature (Richard H. Thaler and Hersh M.
Shefrin 1981).
To address this void in the literature, we incorporate ethics into the traditional principal-
agent model following the two guidelines offered by Koford and Penno (1992). Koford and
Penno suggest that there be a clear boundary between conformity and violation of an ethical
standard, and that ethics be incorporated by adding a penalty for unethical behavior into the
agents utility function. We incorporate the first guideline by allowing the principal to specify a
standard level of effort to the agent at the time of contracting. If the agent accepts the contract,
he is agreeing to provide the specified level of effort in exchange for the compensation. Thereby,
if he fails to provide the effort he violates the ethical norm that valid-agreements-should-be-
kept (Mitnick 1992). We incorporate the second guideline by assuming that the agent suffers a
disutility in the ethical realm when he chooses to provide less than the agreed level of effort. The
magnitude of the disutility depends upon the ethical sensitivity of the agent, which is zero under
the traditional agency assumption of unconstrained opportunism.
We believe that adding a standard for effort and an ethically sensitive agent is a
meaningful extension of the traditional agency model. A standard for effort is already inherent in
the agency literature through the first-best contract and the concept of shirking (Eisenhardt
1989). Even under unobservable effort, the principal knows the effort she would like to induce
from the agent. While the standard becomes irrelevant in traditional agency models, it remains
relevant in our model because it conveys the principals expectation to an ethically sensitive
3

agent. Similarly, we find it reasonable to assume that an ethically sensitive agent would suffer
disutility for violating the standard after agreeing to the contract. This is because individuals tend
to internally monitor themselves and self impose a penalty for behavior that they consider
unethical (Thaler and Shefrin 1981). Moreover, there is a rich literature in ethics examining the
nature of ethical sensitivitywhere it comes from, how it affects behavior, and its impact on
individuals and society. Finally, personal introspection and casual observation persuades us that
experiencing disutility for violating deeply held values such as honesty and duty is as universal
as experiencing utility for wealth or leisure.
We take some level of ethical sensitivity as given, and examine its effect on the
traditional principal-agent model. Our model preserves all other aspects of the traditional agency
framework, including a risk-neutral principal, a risk- and effort-averse agent, unobservability of
effort, and imperfect performance measures. We initially assume that performance measures are
infinitely noisy to make the traditional incentive solution unavailable. This allows us to highlight
the novelty of solutions that emerge when ethics is incorporated into the principal-agent model.
However, after demonstrating the emergence of optimal flat salary solutions, we relax this
assumption and set ethical sensitivity to zero. This allows us to derive the traditional incentive
solution and compare it to the new salary solutions that arise when ethical sensitivity is nonzero.
Introducing a standard for effort and an ethically sensitive agent into the traditional
principal-agent framework generates six main results. The first and fundamental result is the
emergence of a work ethic that controls the opportunism of the agent. We find that the agent
would rather provide the standard level of effort than shirk as long as the standard is below a
critical level and the principal pays him his cost of effort. Intuitively, the critical level of effort
can be interpreted as the maximum effort that the agent perceives to be reasonable or fair.
4

Interestingly, the work ethic imposes an employment ethic because the principal knows that
specifying a standard above the critical level will be perceived as unreasonable by the agent and
will result in shirking. This result embodies the old adage, An honest days pay for an honest
days work.
Our second result is that the principal can induce effort beyond the critical level by
paying the agent a salary that more than compensates him for his cost of effort. This salary
premium offers the agent more than his reservation utility and increases what he considers to be
reasonable effort. Accordingly, the salary premium also imposes an employment ethic in that it
requires the principal to share gains from the agents extra effort.
The third result is that the first-best contract is attainable under unobservable effort for
firms with low productivity relative to the agents ethical sensitivity. This is because the first-
best level of effort is below the critical level for these firms. As such, the principal can obtain the
first-best level of effort by simply requesting it and offering the agent his cost of effort. This is
significant because, while salary contracts are common in practice, they are impossible under
unobservable effort and the traditional assumption of zero ethical sensitivity (unconstrained
opportunism).
3
With zero ethical sensitivity the agents behavior is not influenced at all by the
effort requested by the principal, and financial incentives are necessary to induce any effort from
the agent (Stephen A. Ross 1973, Joel S. Demski and Gerald A. Feltham 1978).
The fourth result of our model is that the optimal salary contract for firms with high
productivity relative to the agents ethical sensitivity is a salary premium contract. For this
category of firms, the first-best level of effort is significantly above the critical level. Thus, it is
optimal for the principal to pay the agent the salary premium and ask for more than the critical
level of effort. In effect, the willingness of the principal to share the incremental benefits
5

motivates the agent to exert the higher effort. The incremental cost, however, makes the optimal
effort lower than the first-best level of effort.
Our fifth result is that the optimal salary contract for firms with medium productivity
relative to the agents ethical sensitivity is an expectation reduction contract. For this category
of firms the first-best level of effort is only marginally above the critical level, and the benefit of
the extra effort does not justify the required salary premium. Accordingly, the principal reduces
her expectation and settles for the critical level of effort. In this case, the work ethic of the agent
restrains the principal from making excessive demands.
The final result of our model relates to comparative statics on ethical sensitivity. We find
that the ethical sensitivity of the agent does not have to reach its upper bound (one) for the first-
best effort to become achievable. In fact, the required level of ethical sensitivity decreases
exponentially with decreases in the firms productivity and the agents risk aversion. Only when
the ethical sensitivity reaches its lower bound (zero), does the flat salary contract lose power to
induce any effort from the agent, making the traditional incentive solution necessary. In general,
however, we show that the traditional incentive solution is more expensive than our flat salary
solution for firms with relatively low and high levels of productivity. Moreover, if the agents
ethical sensitivity is sufficiently high, the flat salary solution dominates the traditional incentive
solution across all firms.
Our results are intuitive and help explain a variety of contracting behavior that is
inconsistent with traditional agency predictions. Take for example the case of not-for-profit
managers. The absence of a bottom-line profit measure makes it difficult to identify suitable
performance measures for incentive contracts. Yet, these managers do not shirk their
responsibility as evidenced by the size and growth of the nonprofit sector in the last two decades.
6

As Susan Rose-Ackerman (1996) points out, this behavior cannot be understood within the
standard agency framework. Among her potential explanations for why non-profit firms exist,
she observes that non-profits recruit ideological managers with high ethical motives relative to
profit motives. This recruiting strategy is confirmed in the financial press (See Max Messmer
2002). By incorporating an ethically sensitive agent, our model is able to explain why non-profit
firms recruit ethical managers and why such managers exert reasonable effort in the absence of
traditional incentives.
Consider also the case of professionals within for-profit firms, such as lawyers, engineers
and accountants. Given that they provide support functions and possess expert knowledge,
bottom line profit measures are not very useful in evaluating their performance and external
monitoring may not be possible (Robert N. Anthony and Vijay Govindarajan 2001). Such
professionals, however, are frequently paid a high flat salary and perform at a high level. The
traditional agency framework has no satisfactory explanation for the behavior of these
professionals either. Our model, however, predicts that where the effort of the agent is highly
valued, the principal will pay a salary premium to get the agent to provide a high level of effort.
Our model is in the spirit of Matthew Rabin (1993), who incorporates fairness into
traditional game theory by adding a preference for fairness into each players utility function.
Rabin uses his model to explain experimental results from cooperation games (e.g., public goods,
prisoners dilemma) suggesting that people cooperate to a greater degree than would be implied
by pure self-interest. Similar to our ethics results, Rabin finds that agents trade off their
preference for fairness against their preference for earnings, and one preference may dominate
the other in a given situation. In a special application of his model, Rabin considers a situation in
which a worker chooses an effort level (High or Low) and the firm simultaneously chooses a flat
7

salary for the worker that shares a portion of the firms income. Similar to our results, Rabin
finds that flat salary contracting is not possible without some preference for fairness. Our model
is unique, however, in that it incorporates a preference for ethics within the traditional principal-
agent framework. Thus, we are able to provide new and useful insights to augment the extensive
agency literature in accounting, economics, and finance.
The rest of the paper is organized as follows. In the following section we present our
model more formally and demonstrate how we introduce a standard for effort and an ethically
sensitive agent. In section II we present the main results of the model. We compare our results to
the traditional agency model in section III and discuss the implications and directions for future
research in section IV. We conclude in section V.
I. THE MODEL
We begin with the basic single-period, principal-agent framework. The principal is risk
neutral
4
and hires a risk- and effort-averse agent to perform a production task that she cannot do
herself. The outcome the principal expects to realize at the end of the period is a multiple of the
firms productivity p and the agents effort a:
(1) | | pa Y = E
A moral hazard problem arises because the agent experiences disutility from effort and the
principal is unable to monitor the agents effort. The traditional agency solution to this problem
is to provide incentives to the agent based on a performance measure X that serves as a noisy
signal of the agents effort:
5

(2) X a = + ; where ( )
2
0, N ~ .
8

The noise term prevents the principal from perfectly inferring the agents effort, and
represents other unobservable factors that affect the performance measure X. This noise term is
assumed to be normally distributed with mean zero and variance
2
. The variance captures the
level of inefficiency in X as an inference tool. We initially assume an infinitely large variance,
characterizing a situation where the performance measure is not a reliable signal of the agents
effort. Thus, the principal can only utilize a flat salary contract.
Figure 1 presents the events in our model. As is true of most agency models, there is no
asymmetry of pre-contract information (Baiman 1982). That is, both the principal and the agent
know the production technology of the firm,
6
the characteristics of the performance measure, and
the preferences of each other (including the ethical sensitivity of the agent) before contracting
takes place. At the time of contracting, however, we allow the principal to specify a standard
level of effort, d, in addition to the compensation wage, . Thus, if the agent accepts the
contract, he is explicitly agreeing to exert the standard level of effort in return for the specified
wage, although his choice of effort a is not observable by the principal. If instead the agent
rejects the contract, he receives his reservation utility from the labor market.
[Insert Figure 1 about here]
The economic concepts of opportunism and shirking are based upon some previously
agreed-upon level of effort (Eisenhardt 1989), so a standard for effort is inherent in agency
theory. In the traditional case, however, the existence of the standard is irrelevant because the
agent is opportunistic and suffers no disutility when choosing to deviate from it. In traditional
agency models, therefore, the standard is not an argument in the model even though the principal
knows the effort she would like to induce. Instead, the principal indirectly induces effort from
9

the agent through an incentive contract. In our model, however, the standard specified by the
principal is relevant because it raises a moral obligation in an ethically sensitive agent.
We incorporate ethical sensitivity into the model in the following manner. If the agent
accepts the contract, he agrees to provide the standard level of effort d. Therefore, if he provides
a level of effort less than d, he is going against his word and failing to fulfill his obligation to the
principal. We use an indicator variable to capture the agents violation of the standard.
Specifically, is one when the agent violates the standard (a < d) and zero when the agent meets
his obligation under the contract (a d). Moreover, we assume that the agent suffers a utility
loss e when he chooses to violate the standard for effort after agreeing to the contract. Thus, we
incorporate a multiplicative term, e, to the standard utility function to capture the potential
disutility from violating the standard for effort:
7

(3) ( ) ( ) e a a = , U , , U
where ( ) U , U a and 0 e U
We allow the utility loss e associated with failing to meet the standard to vary from 0 to
U, the upper bound of the agents utility from the physical realm. Note that e = 0 corresponds to
the traditional agent who is fully opportunistic and insensitive to the standard specified by the
principal, and e = U corresponds to an agent who will never shirk once he agrees to the contract.
Accordingly, e corresponds directly to the agents ethical sensitivity, i.e., the depth of his
conviction for values such as honesty and duty.
Ethical sensitivity has a variety of potential sources. It could be inborn or genetic (Jack
Hirshleifer 1977 and Gary S. Becker 1976), but is more likely to arise from childhood and school
socialization (Koford and Penno 1992). Religion has also served an important role in
10

establishing and maintaining ethical codes and convictions (Noreen 1988). Within the
organization, ethical values can also be attributed to management leadership, company policies,
and government regulations (David J. Cherrington 1980). Although ethical sensitivity is
exogenous to our model, we assume that it varies across agents due to differences in upbringing,
socialization and training. In fact, a potential theme from our model is the considerable benefit of
encouraging ethical sensitivity, and we discuss this at length in a later section.
Our assumption that ethical sensitivity is exclusively in the ethical realm is clearly
reflected in equation 3, as the agents utility loss from violating the standard is independent of
the level of pay and effort. We add to this assumption the standard assumptions contained in
traditional LEN agency models (e.g., Bengt Holmstrom and Paul Milgrom 1991, Feltham and
Xie 1994, Datar, Kulp, and Lambert 2001).
8
In particular, we assume that utility is strictly
decreasing in effort (

<
u
a
0) and increasing in income (

>
u

0), and that the agent has constant


absolute risk aversion (r) and finds his work less onerous as his income increases (


>
2
0
u
a
).
The later assumption is present in all LEN agency models, possibly because researchers have
found it reasonable to assume that the agent will be willing to exert more effort if he is given a
higher wage. This willingness to exert more effort could come from the agents increased sense
of prestige in his work, or his increased sense of gratitude and loyalty to the firm because he is
unlikely to get the same high pay from the external labor market.
In keeping with our intent to incorporate ethics into the traditional LEN agency
framework, we utilize the following exponential form for the agents utility:
9

(4) ( ) ( )
( )
2
2
U , , U , 1 exp
a
a a e r e
(
= =


where ( ) U , a =1 and 0 1 e
11

The principals problem in our setting is to specify a wage and a standard for effort that
will maximize her expected utility subject to two constraints. This is formally stated as:
(5)
Max:
, d
( ) E pa
Subject to:
IR
( )
{ }
2
2
E 1 exp 0
a
r e
(



IC
( )
{ }
2
2
,
, Arg max E 1 exp
a
a
a r

e
(



The first constraint represents the assumption of individual rationality (IR), and ensures
that the agent will only contract with the principal if it is in his best interest. Without loss of
generality, we have assumed that the agents reservation utility is zero.
10
The second constraint
represents the assumption of incentive compatibility (IC), and ensures that the agent will
maximize his utility in choosing the level of effort and, correspondingly, whether or not to meet
the standard. If the agent chooses to meet the standard, = 0 and he provides effort a = d. But if
the agent chooses not to meet the standard, = 1 and he will provide a level of effort a below d
that maximizes his expected utility. Where only flat salary contracts are available, the level of
effort that maximizes the agents utility is zero once he chooses not to meet the standard.
Note that setting e = 0 makes an irrelevant variable for the IR and IC constraints. In this
case, the principal derives no benefit from specifying a desired level of effort because the agent
suffers no disutility from shirking. This emphasizes again that setting the agents ethical
sensitivity to zero gives us the traditional agency case, which is void of any ethical
considerations.
12

II. MAIN RESULTS
We begin our analysis by deriving the first-best contract. Here we assume that the
principal can costlessly monitor the agents effort and inflict a severe penalty if the agent does
not provide the standard level of effort. In this setting, the agent is compelled to provide the
standard level of effort upon contracting ( = 0), and the IC constraint is redundant. Substituting
=
a
2
2
from the IR constraint into the principals objective function in equation 5 results in:
(6)
Max:
a
pa
a

2
2
.
Note that the term pa captures the principals benefit and the term
a
captures the principals
cost of inducing effort a. The first-order condition of this problem results in:
2
2
(7) a p
FB
= ,
where a
FB
is the first-best effort and is increasing in the firms productivity p. Interestingly,
introducing a standard for effort and an ethically sensitive agent has not altered the first-best
contract from the extant agency literature. It remains a flat-wage contract that pays the agent his
cost of providing the first-best effort:
(8)
FB
p
=
2
2
.
The agents ethical sensitivity plays no role in the first-best solution because the threat of a
severe penalty (external monitoring) makes the utility loss from violating the standard (internal
monitoring) redundant. The first-best solution is presented graphically in Figure 2.
[Insert Figure 2 about here]
13

When effort is unobservable, however, our results diverge widely from the traditional
principal-agent model. Given traditional agency assumptions, a flat salary contract has no power
to induce effort from the agent in this case (Ross 1973, Demski and Feltham 1978). This changes
when we introduce ethical sensitivity to the model. Specifically, a work ethic emerges that
deters the opportunism of the agent and grants a salary contract the power to induce effort. This
result is stated formally in our first proposition.
PROPOSITION 1: The principal can induce a level of effort d that is below a critical level d
1

by simply requesting it from the agent and paying him a flat salary that equals the cost of that
effort,
(9)
2
2
d
= for
1
2 1
1
log
r
e
d d

= ,
where e and r are the ethical sensitivity and risk aversion of the agent respectively. If the
requested level of effort is above this critical level, however, the agent will shirk completely and
exert no effort (i.e., a = 0).
Proof: All proofs are in an appendix
The intuition behind Proposition 1 is not difficult to see. Upon contracting with the
principal, the agent has two options. He can exert the standard effort and thereby avoid the utility
loss e, or he can violate the standard and provide zero effort. In the latter case, he incurs the
utility loss e but avoids all disutility for effort. Yet, when the standard level of effort is below the
critical level, the agent would rather provide the standard than shirk because the gain from
avoiding the disutility for effort is not large enough to justify the utility loss in the ethical realm.
Intuitively, the critical level of effort d
1
represents the maximum level of effort that the
agent finds reasonable or fair. Equation 9 suggests that the critical level of effort is determined
14

by the agents ethical sensitivity and risk aversion. In particular, d
1
is increasing in ethical
sensitivity and decreasing in risk aversion. When the agent has zero ethical sensitivity, d
1
is zero
and the flat wage is incapable of inducing any effort from the agent as in the traditional case. As
ethical sensitivity increases above zero, however, the agent finds a higher level of effort to be
reasonable.
The result that d
1
is decreasing in r may seem surprising given that the agent bears no
financial risk with a flat salary contract. But r plays an important role in our model by
influencing the agents tradeoff between the physical realm and the ethical realm. The greater the
agents risk aversion, the more he values the physical realm relative to the ethical realm.
Therefore, the agent appreciates more the increase in net wealth that comes from shirking (i.e.,
saving on effort). In essence, an increase in risk aversion reduces the effect of ethical sensitivity
by making the ethical realm relatively less important to the agent. This ethics dampening role
of risk aversion is ignored in traditional agency models that assume zero ethical sensitivity.
It is important to distinguish between ethical sensitivity and the emerging work ethic in
our model. As discussed above, ethical sensitivity captures the depth of the agents ethical values
and is exogenous to the model. In contrast, the work ethic captures the agents willingness to
exert effort, which arises endogenously from our model as a result of the agents ethical
sensitivity and the standard for effort. It is also important to note that the work ethic does not
reduce the agents disutility for effort.
11
Rather, the work ethic reflects the amount of effort that
the agent is willing to provide despite his disutility because of his sensitivity to ethical
considerations. Thus, the emerging work ethic in our model is consistent with much of the work
ethic literature (See Cherrington 1980 and Irving H. Siegel 1983).
15

Interestingly, Proposition 1 suggests that the agents ethical sensitivity also imposes an
employment ethic on the principal. This is because the principal cannot specify a standard for
effort higher than the critical level, or the agent will shirk. Specifically, the principal will never
ask for a level of effort d greater than d
1
for a flat wage of =
d
2
2
. Yet, a second result of our
model is that the principal can induce a level of effort greater than the critical level (d > d
1
) by
offering the agent a salary premium. This result is stated formally in Proposition 2:
PROPOSITION 2: The principal can induce a level of effort d that is greater than the critical
level d
1
by paying the agent a flat salary that includes a salary premium,
(10)
2
2
2
1
2
1 exp
log
rd
e
d
r

| |

\

|
.
(
( =
(

+ for d d .
1
Proof: See appendix
To see the intuition behind Proposition 2, recall that an increase in income makes effort
less onerous to the agent and thereby decreases the gain from shirking. At the salary premium
described in Proposition 2, the gain from shirking exactly equals the loss in the ethical realm.
Thus, paying the agent the salary premium in addition to his cost of effort makes him willing to
exert a level of effort d above the critical level d
1
. This result suggests, however, that to induce
effort beyond the critical level the principal must share the resulting increase in profit with the
agent. We graph the cost of effort and the salary premium from equation 10 in Figure 3.
[Insert Figure 3 about here]
Figure 3 shows that the cost of the salary premium is concave in d. Initially, the principal
must be willing to share more and more of the increase in profit with the agent to induce
incremental increases in effort. Thereafter, the agent requires less and less to induce effort and
16

the salary premium converges to an upper bound equaling
1
log
r
1
e
as d goes to infinity.
Interestingly, the cost of the salary premium decreases with the ethical sensitivity of the agent e.
This is because the gain from shirking must exceed a higher threshold, making it less attractive.
Moreover, d
1
increases with e, thereby reducing the amount of effort the principal needs to
induce through the salary premium.
Now that we have established that the principal can induce effort from an ethically
sensitive agent through a flat salary, we turn our attention to the optimal flat salary contract. We
first note that at the optimal flat salary, the induced level of effort a always equals the standard
level of effort asked by the principal, d. This is because if a < d, the agent suffers the ethical
disutility e and subsequently has no incentive to exert any positive level of effort. On the other
hand, if a > d the contract will not be incentive compatible as the agent can improve his utility by
reducing his effort to d. Thus, the principals problem reduces to finding the d that maximizes
her expected outcome, , net of the wage function | | pd Y = E d a f. We continue our analysis by
deriving the optimal salary contract for firms with different levels of productivity, p.
PROPOSITION 3: The optimal flat salary for firms with relatively low productivity, i.e.,
1
2 1
1
log
r
e
p d

= , is the first-best contract. For this category of firms, the principal will
specify a standard level of effort that is equal to the first-best effort, d d
1 FB
, and pay a flat
salary equal to the cost of that effort,
2
2
p
= .
d = <
Proof: See appendix
Proposition 3 shows that for low productivity firms, the principal will ask the agent for
the first-best level of effort in exchange for a salary that pays the cost of that effort, and the agent
17

will provide it. This is because the first-best level of effort for these firms is below the critical
level. Thus, the first-best solution is achievable for this category of firms even under
unobservable effort.
Next, we derive the optimal contract for firms with relatively high productivity.
PROPOSITION 4: The optimal flat salary for firms with relatively high productivity, i.e.,
2
2
2
2
1 exp
rd
p
|
|
|
\ .

>
d
|
1
where d , is a salary premium contract. d
2

12
For this category of firms, the
principal will specify a standard level of effort that is between the critical level and the first-best
effort,
( ) 1 2
2
FB
rd
d d d p d < = <
2
1 exp

, and pay a flat salary that includes the cost of that


effort and a salary premium,
2
2
2
1
2
log
rd
d
r
| |
|
\ .
1 exp
e

(
( =
(

+ .
Proof: See appendix
Proposition 4 shows that for high productivity firms, the principal will induce the agent
to exert more than the critical level by paying the salary premium. This is because the first-best
level of effort is sufficiently above the critical level and the increase in profit justifies the
increased cost of the salary premium. However, the incremental cost of the salary premium
makes the optimal level of effort lower than the first-best level.
Next, we derive the optimal contract for firms with medium productivity.
PROPOSITION 5: The optimal flat salary for firms with medium productivity, i.e.,
2
2
1
2
2
1 exp
rd
d p
|
|
|
\ .


d
|
d
1
where d , is an expectation reduction contract. For this category of
2

18

firms, the principal will specify a standard level of effort that is equal to the critical level,
, and pay a flat salary that equals the cost of that effort, d d d
FB
= <
1
2
1
2
d
= .
Proof: See appendix
Proposition 5 shows that for medium productivity firms, the principal will reduce her
expectations and ask the agent for the critical level of effort in exchange for the cost of that
effort, and the agent will provide it. This is because, while the first-best level of effort for these
firms is above the critical level, the potential increase in profits is not sufficient to justify the
added cost of the salary premium.
We conclude our analysis by performing comparative statics on ethical sensitivity. In
particular, we determine the level of ethical sensitivity above where the optimal flat salary
contract is capable of inducing the first-best effort, and the level of sensitivity below where the
salary contract has zero power to induce effort.
PROPOSITION 6: The optimal flat salary contract induces the first-best effort if
(11)
2
2
1 exp
p r
e
| |

|
\ .
,
and is incapable of inducing any effort if and only if e = 0.
Proof: See appendix
Proposition 6 reveals that the ethical sensitivity of the agent does not have to approach its
upper bound (one) for the optimal flat salary contract to become capable of inducing the first-
best effort. As equation 11 shows, the ethical sensitivity required to achieve first-best decreases
exponentially with decreases in risk aversion and productivity. As such, the required ethical
sensitivity can be relatively low for agents with low risk aversion and firms with low
19

productivity. Proposition 6 also reveals that the flat salary contract loses the power to induce any
effort from the agent if and only if the ethical sensitivity takes the extreme value of zero. Taken
together, this suggests that ethical sensitivity is a very efficient motivator that should not be
ignored by agency researchers.
III. COMPARISION TO THE TRADITIONAL AGENCY SOLUTION
In this section we compare the results of our model to the traditional agency model by
contrasting our solutions with the traditional incentive solution. First, we derive the traditional
incentive solution by relaxing the assumption that the noisiness
2
of the performance measure is
infinitely large and imposing the assumption of zero ethical sensitivity. Consistent with the
traditional LEN framework discussed earlier, we assume the employment contract is linear in the
performance measure ( X B A + = ) in addition to our assumptions that the agents utility is
exponential and the performance measure is normally distributed. In this setting the principals
problem is to specify A and B that will maximize her expected utility subject to the incentive
compatibility (IC) and individual rationality (IR) constraints. In the appendix we present the
traditional agency problem and show that equation (5) reduces to
(12)
Max:
a
| |
E pa
Subject to:
IC B a =
IR
| |
( )
2
2 2
2 2
E
a a
r = +
20

The IC constraint suggests that the effort the principal can induce is directly proportional
to the incentive rate B. Thus, a flat salary ( ) is incapable of inducing any effort from the
agent. The IC constraint also suggests that A plays no role in the optimal level of effort. Thus,
the principal will set A in such a way that there is no slack in the IR constraint, which allows us
to derive an expression for the expected cost of the compensation. According to the IR
constraint, the risk averse agent will require not only his cost of effort (
a
2
2
) from the principal,
but also a risk premium (
2
2
a
2
r ) to compensate him for carrying the risk. This risk premium is
increasing in the agents risk aversion r and the noisiness
2
of the performance measure.
0 B =
Thus, the optimal level of effort a that the principal will induce in this setting is:
(13)
Max:
a
( )
2
2 2
2 2
a a
pa r
(


+
with the resulting solution,
(14)
2
1
I
p
a
r
=
+
.
Equation 14 shows that the optimal level of effort in this setting is clearly below the first best
level, p (see equation 7). Further, the optimal level of effort falls further and further below first-
best with increases in r or
2
. We graph the cost of effort and the risk premium from equation 13
in Figure 4.
[Insert Figure 4 about here]
Next, we turn our attention to contrasting the traditional incentive solution to the flat
salary solutions that arise with ethical sensitivity in the model. Note that the principals benefit
from the agents effort (pa) is the same under both schemes. Similarly, under both schemes the
principal must pay the agent his cost of effort (
a
2
2
). Therefore, the substantive difference in the
21

two solutions is between the salary premium that arises in our model versus the risk premium in
the traditional model.
Figure 5 compares the concave salary premium in our model with the traditional convex
risk premium. First of all, note that as the noisiness of the performance measure approaches
infinity (i.e., ), as we initially assumed, the risk premium curve rises to the vertical axis
and is universally dominated by the salary premium. On the other hand, as ethical sensitivity
approaches the traditionally assumed value of zero (i.e., ), the salary premium curve rises
toward the vertical axis and is universally dominated by the incentive premium.

2

0 e
[Insert Figure 5 about here]
The more interesting setting, however, is where ethical sensitivity is nonzero and the
noisiness of the performance measure is finite. Figure 5 suggests that for moderate levels of
ethical sensitivity and noise, the salary premium will be cheaper than the traditional risk
premium when the effort to be induced (or productivity) is either low (Region I) or high (Region
III). In contrast, when the effort to be induced is at intermediate levels (Region II), the risk
premium is cheaper than the salary premium. This suggests that for firms with intermediate
productivity, the principal will incorporate financial incentives instead of implementing an
expectation reduction contract as in our proposition 5. Yet, as ethical sensitivity increases, the
salary premium curve moves down and toward the right and eventually dominates the risk
premium curve for every firm. On the other hand, irrespective of how small ethical sensitivity is,
the risk premium will never universally dominate the salary premium solution. These
comparisons suggest that agency researchers have overlooked an important aspect of the
principal-agent relationship by ignoring ethical sensitivity. We leave it to future research to solve
22

for the optimal mix of flat salary and financial incentives in the presence of an ethically sensitive
agent.
IV. IMPLICATIONS AND DIRECTIONS FOR FUTURE RESEARCH
We discuss the implications of our model for various streams of economic research and
provide directions for future research. Specifically, we address the implications of our model for
research in work ethics, efficiency-wage theory, reputation, and executive compensation.
Our model has direct implications for the work ethic literature. Research in this area
examines observed or implied values regarding work, the forces that affect these values, and the
effects of these values on economic outcomes (See Cherrington 1980). In their review of the
literature, Paul J. Andrisani and Herbert S. Parnes (1983) conclude that the strength of an
individuals work ethic affects their success in the labor market.
13
They assert that such findings
have significant implications for public policy.
14
Our model provides formal support for these
findings. We find that a work ethic emerges from the agents deeply held values for honesty and
duty. This work ethic increases the welfare of the agent by generating employment opportunities
that would otherwise be impossible, and by allowing him to receive a salary that pays more than
his reservation utility if his effort is highly valued. This work ethic also benefits the principal by
reducing the cost of inducing effort from the agent. This suggests that the emphasis parents place
on values such as honesty and duty not only benefits their children, but the economy and society
at large.
15

Our model also addresses the concerns of researchers who view the traditional work ethic
as an attempt by capital to extort more work and profit from labor (See Gus Tyler 1983). In our
model, the agents work ethic generates an employment ethic that controls the opportunism of
23

the principal. In particular, the principal cannot induce more than a critical level of effort without
paying the agent a salary premium that shares the increase in profit with the agent. In contrast, in
the absence of ethical sensitivity and the emerging work ethic, the principal is forced to utilize an
incentive contract that inflicts financial risk and only grants the agent his reservation utility.
Finally, our model shows that an ethically sensitive agent will directly benefit from productivity
gains that arise due to improvements in the firms production technology. As productivity
increases, our model predicts that the agent will extract a higher share of the incremental profits
from the principal.
Our model also has implications for efficiency-wage theory. According to efficiency-
wage theory, firms pay more than the market-clearing wage to extract additional effort from the
agent (George A. Akerlof 1984). This wage premium functions similar to our salary premium,
but the mechanism is very different. In efficiency-wage theory, a higher wage discourages
shirking because of the possibility of detection followed by possible unemployment. In contrast,
our model attributes the salary premium to a principal attempting to induce more than a critical
level of effort from an ethically sensitive agent. Our results also suggest that salary premiums
will only appear where the productivity of the firm is high relative to the ethical sensitivity of the
agent. This prediction is consistent with empirical observations that salary premiums appear only
in high productivity industries (Thaler 1989).
Because our model is a single-period model and the principal knows the ethical
sensitivity of the agent, concerns for reputation do not endogenously arise. However, our model
demonstrates that ethical sensitivity is able to control opportunism quite apart from reputation
effects. This is consistent with experimental results in Stevens (2002) that ethical concerns
caused subjects to reduce their budgetary slack even when anonymity was carefully preserved.
24

Moreover, by showing that ethical sensitivity benefits both the principal and the agent, our
model provides a rationale for agents to develop and maintain their ethical reputation over time.
The value of ethical sensitivity is supported by the significant resources firms spend on ethical
screening and training and recruits and employees spend on signaling their ethical type through
reference letters and affiliations. This suggests that ethical sensitivity may actually increase in
significance in a multiple-period setting. Thus, an obvious extension of our model is to
incorporate a multiple-period setting where the agents ethical sensitivity is unobservable to
study the interplay between ethical sensitivity and reputation effects.
Finally, our model has implications for executive pay. While recent scandals have
focused on the seemingly excessive compensation of top executives in U.S. firms, our model
predicts the existence of such compensation schemes where the executives effort is highly
valued (high productivity) relative to their ethical sensitivity. This is a somewhat surprising
result. That is, the inclusion of ethical sensitivity points to the existence of potentially high flat
salaries for highly productive agents with relatively low ethical sensitivity. In fact, our model
would suggest that the dramatic increase in executive salary pay over the past few decades is
indicative of a general decrease in ethical sensitivity.
There are a number of other potential extensions to our research. As mentioned
previously, an important extension of our study is to examine the role of financial incentives in
the presence of an ethically sensitive agent. Future research is needed to examine how the
emerging work ethic and salary premium solutions documented here affect the extant incentive
solutions from traditional agency models. This is a particularly interesting exercise given the
additional role of risk aversion (ethics dampening) that comes into play with the introduction of
ethical sensitivity. This research could also pave the way for research regarding the implications
25

of incentive mechanisms for the ethical sensitivity of the agent. Some empirical evidence
suggests that such incentives may actually reduce the ethical sensitivity of the agent (Geoffrey
Sprinkle 2003). For example, the recent emphasis on incentives may have increased the tendency
for executives to behave opportunistically and game the system.
Future research could also expand the functional form of our ethical disutility for
violating the standard for effort. We have assumed that the agent experiences the same ethical
disutility for all effort below the standard. In reality, the ethical disutility may only begin at some
level below the standard and/or increase as the gap between the effort and the standard grows.
That is, the agent may justify to himself a minimum level of deviation from the standard. The
results of our model, however, appear robust to such changes in the functional form of the ethical
disutility. Future studies could also examine other potential unethical behavior by the agent, or
examine unethical behavior on the part of the principal. Finally, further research is needed to
guide our understanding of ethical decision-making when multiple tasks are present.
Incorporating multiple tasks would allow the model to examine the trade-off between some tasks
that are ethical and some that are unethical (e.g., earnings manipulation).
V. CONCLUSION
We examine the common assertion that incorporating ethics would enhance the
descriptive validity of agency theory (Arrow 1985, Noreen 1988, Mitnick 1992, Koford and
Penno 1992). We add ethics to the principal-agent model in a rigorous and parsimonious way,
and preserve all other aspects of the traditional agency framework. In our model, the principal
specifies a standard for effort at the time of contracting and the agent suffers a utility loss if he
chooses not to provide the standard after agreeing to the contract. The magnitude of the loss
26

depends upon the agents ethical sensitivity, which captures the depth of his values for honesty
and duty. A standard for effort is already inherent in the agency literature through the first-best
contract and the concept of shirking (Eisenhardt 1989), and we believe that an agent would likely
suffer disutility for going against his word and failing to meet his obligation to the principal.
Thus, our model captures the inherent trade-off between the physical realm and the ethical realm
that appears from the inception of economic theory (Smith 1759) and has been documented in
the experimental literature.
Incorporating ethics into the model in this way causes optimal salary contracts to arise
that are impossible in the traditional agency case. We argue that these contracts are descriptive of
employment contracts found in practice. For example, our model explains the effectiveness of
salary contracts in the not-for-profit sector and within highly skilled professions where
performance measures for incentive contracting are very difficult to find. Further, our salary
premium solution helps explain the existence of high salary contracts for workers and top
executives whose effort is highly valued relative to their ethical sensitivity. When we compare
our salary solution to the traditional incentive solution, we find that the traditional incentive
solution is more expensive than the flat salary solution for firms with relatively low and high
levels of productivity. Further, for agents with high levels of ethical sensitivity, we show that the
salary solution dominates the traditional incentive solution across all firms. Thus, our results
support Arrows (1985) assertion that organizational control is likely to rely heavily on informal
expectations and systems of ethics.
Our model also confirms the common assertion that introducing ethics into the traditional
agency model would enhance the pedagogical usefulness of the theory. Researchers have noted
the obvious inconsistency of emphasizing the importance of professional ethics while basing
27

business research and teaching on models of human behavior that are totally void of ethical
content (Noreen 1988, Wanda A. Wallace 1992). Our model demonstrates that ethical agents are
of great value to the firm, and provides a rationale for firms to hire ethical agents and provide
ethics training. Further, our model demonstrates the significant loss of efficiency that can occur
when ethical sensitivity is allowed to diminish. As such, ethical sensitivity is an essential human
trait that is worth cultivating in our homes, communities, and business schools.
28

REFERENCES
Akerlof, George A. Loyalty Filters. American Economic Review, March 1983, 73(1), pp. 54-
63.
________. Gift Exchange and Efficiency-Wage Theory: Four Views. American Economic
Review Proceedings, 1984, 74(2), pp. 79-83.
Andrisani, Paul J. and Herbert. S. Parnes. Commitment to the Work Ethic and Success in the
Labor Market: A Review of Research Findings, in Jack Barbash, Robert J. Lampman,
Sar A. Levitan and Gus Tyler, eds., The Work EthicA Critical Analysis. Industrial
Relations Research Association. Madison, WI, 1983, pp. 101-120.
Anthony, Robert N. and Vijay Govindarajan. Management Control Systems. 10th ed. New
York: McGraw-Hill Irwin, 2001.
Arrow, Kenneth J. The Economics of Agency, in John W. Pratt and Richard J. Zeckhauser,
eds., Principals and Agents: The Structure of Business. Boston, MA: Harvard Business
School Press, 1985, pp. 37-51.
Baiman, Stanley. Agency Research in Managerial Accounting: A Survey. Journal of
Accounting Literature, 1982, 1, pp. 154-210.
________. Agency Research in Managerial Accounting: A Second Look. Accounting,
Organizations, and Society, 1990, 15(4), pp. 341-71.
Banfield, Edward C. The Moral Basis of a Backward Society. The Free Press: New York. 1958.
Becker, Gary S. Altruism, Egoism, and Genetic Fitness: Economics and Sociobiology.
Journal of Economic Literature, September 1976, pp. 817-26.
Besley, Timothy and Maitreesh Ghatak. Competition and Incentives with Motivated
Agents. Working Paper, London School of Economics, 2003.
29

Cherrington, David J. The Work Ethic: Working Values and Values That Work. New York:
AMACOM, 1980.
Datar, Srikant, Susan Cohen Kulp, and Richard A. Lambert. Balancing Performance
Measures. Journal of Accounting Research, June 2001, 39(1), pp. 75-92.
Demski, Joel S. and Gerald A. Feltham. Economic Incentives in Budgetary Control
Systems. The Accounting Review, 1978, 53(2), pp. 336-59.
Eisenhardt, Kathleen M. Agency Theory: An Assessment and Review. Academy of
Management Review, 1989, 14(1), pp. 57-74.
Evans, J. Harry III, R. Lynn Hannan, Ranjani Krishnan, and Donald V. Moser. Honesty
in Managerial Reporting. The Accounting Review, October 2001, 76, pp. 537-59.
Feltham, Gerald A. and Jim Xie. Performance Measure Congruity and Diversity in Multi-
Task Principal/Agent Relations. The Accounting Review, July 1994, 69(3): 429-53.
Hirshleifer, Jack. Economics from a Biological Viewpoint. The Journal of Law and
Economics, April 1977, pp. 1-52.
Holmstrom, Bengt and Paul Milgrom. Multitask Principal-Agent Analyses: Incentive
Contracts, Asset Ownership and Job Design. Journal of Law, Economics and
Organization, 1991, 7, pp. 24-51.
Koford, Kenneth and Mark Penno. Accounting, Principal-Agent Theory, and Self-Interested
Behavior, in Norman E. Bowie and R. Edward Freeman, eds., Ethics and Agency
Theory: An Introduction. New York, NY: Oxford University Press, 1992, pp. 127-42.
Messmer, Max. Launching a Career in the Nonprofit Sector. Strategic Finance, November
2002, 84(5), pp. 15-16.
30

Mitnick, Barry M. The Theory of Agency and Organizational Analysis, in Norman E. Bowie
and R. Edward Freeman, eds., Ethics and Agency Theory: An Introduction. New York,
NY: Oxford University Press, 1992, pp. 75-96
Noe, Thomas H. and Michael J. Rebello. The Dynamics of Business Ethics and Economic
Activity. The American Economic Review, June 1994, 84(3), pp. 531-47.
Noreen, Eric W. The Economics of Ethics: A New Perspective on Agency Theory.
Accounting, Organizations, and Society, 1988, 13(4), pp. 359-69.
Pratt, John W. and Richard J. Zeckhauser. Principals and Agents: An Overview, in John
W. Pratt and Richard J. Zeckhauser, eds., Principals and Agents: The Structure of
Business. Boston, MA: Harvard Business School Press, 1985, pp. 1-35.
Putnam, Robert D. Democracies in Flux: The Evolution of Social Capital in Contemporary
Society. Robert D. Putnam, ed. Oxford University Press: Oxford. 2002.
Rabin, Matthew. Incorporating Fairness into Game Theory and Economics. American
Economic Review, December 1993, 83(5), pp. 1281-302.
Rose-Ackerman, Susan. Altruism, Non-profits, and Economic Theory. Journal of Economic
Literature, June 1996, 34(2), pp. 701-728.
Ross, Stephen A. The Economic Theory of Agency: The Principals Problem. The American
Economic Review, May 1973, 63(2), pp. 134-9.
Schatzberg, Jeffrey W. and Douglas E. Stevens. Budgetary Slack and Shirking in
Participative Budgeting: An Experimental Investigation of Opportunism, Fairness, and
Ethics. Working Paper, Syracuse University, 2003.
31

Siegel, Irving. H. Work Ethic and Productivity, in Jack Barbash, Robert J. Lampman, Sar A.
Levitan and Gus Tyler, eds., The Work EthicA Critical Analysis. Industrial Relations
Research Association. Madison, WI, 1983, pp. 27-42.
Simon, Herbert A. Organizations and Markets. Journal of Economic Perspectives, Spring
1991, 5(2), pp. 25-44.
Smith, Adam. Theory of Moral Sentiments. London: Millar, 1759.
Sprinkle, Geoffrey B. Perspectives on Experimental Research in Managerial Accounting.
Accounting, Organizations and Society, 2003, 28, pp. 287-318.
Stevens, Douglas E. The Effects of Reputation and Ethics on Budgetary Slack. Journal of
Management Accounting Research, 2002, 14, pp. 153-171.
Stiglitz, Joseph E. Symposium on Organizations and Economics. Journal of Economic
Perspectives, Spring 1991, 5(2), pp. 15-24.
Thaler, Richard H. Interindustry Wage Differentials. Journal of Economic Perspectives,
1989, 3(2), pp. 181-93.
________ and Hersh M. Shefrin. An Economic Theory of Self-Control. Journal of Political
Economy, Spring 1981, 89(2), pp. 392-406.
Tyler, Gus. The Work Ethic: A Union View, in Jack Barbash, Robert J. Lampman, Sar A.
Levitan and Gus Tyler, eds., The Work EthicA Critical Analysis. Industrial Relations
Research Association. Madison, WI, 1983, pp. 197-210.
Wallace, Wanda A. Integrating Ethics into Doctoral Education: The Apparent Dilemma of
Agency Theory, in Norman E. Bowie and R. Edward Freeman, eds., Ethics and Agency
Theory: An Introduction. New York, NY: Oxford University Press, 1992, pp. 127-42.
32

ENDNOTES
1. Koford and Penno (1992) consider two alternative ways of modeling ethical behavior: (1)
some people are ethical and others are not, and (2) any individual will be ethical in some
situations and not in others. In the first case, they show that the cost of internal control
systems decreases as the proportion of ethical agents increases, and therefore attracting
ethical agents benefits the firm. In the second case, they show that an accounting system can
benefit the firm by clarifying and enforcing ethical boundaries.
2. Mitnick (1992, 80) describes principal-agent models as careful and sometimes elegant stuff,
even though what is established is usually constrained by unrealistic assumptions. He
challenges modelers to keep testing the bounds of formalism; at some point the marginal
advances may add up or the techniques advance closer to realism.
3. Bengt Homstrom and Paul Milgrom (1991) demonstrate that when there are multiple tasks,
there exist special situations where the optimal contract will be a flat wage. It is also known
that flat salaries can induce effort if there are implicit incentives such as fear of dismissal or
concern for reputation (Baiman 1990). Such incentives do not exist, however, within single-
period models.
4. The argument behind a risk neutral principal is that she is capable of well diversifying her
investments.
5. The signal of the agents effort could be a multiple of the outcome Y itself (i.e.,
Y
p
X = ).
6. In fact, our results do not require that the agent know the production technology of the firm
reflected in equation 1.
7. We recognize that the introduction of a standard may bring about other intangibles that may
be included in the agents utility function. For example, the agent may feel a sense of
33

personal satisfaction for exceeding the standard. We consider this a secondary effect and
focus on the impact of conformity versus violation of a standard for effort. Similarly, agents
may not suffer the full ethical disutility e until their effort is somewhat below the standard.
However, we expect our results to hold even if we model the ethical disutility to reflect such
characteristics.
8. Traditional LEN agency models assume a Linear incentive contract, Exponential utility for
the agent, and Normal distribution of noise terms. Our model is consistent with the LEN
framework contained in these models.
9. The popularity of this exponential form in the incentive literature arises from its power to
produce closed form solutions to the principals problem when it is combined with normally
distributed performance measures (X for example) and linear contracts ( X B A + = ).
Accordingly, assuming this form allows us to compare and contrast our optimal flat salary
contracts with the traditional incentive contract.
10. The assumption of a zero reservation wage is mainly for computational convenience. Our
results hold if some nonzero reservation wage is assumed.
11. While there may be aspects of work that are valued by the agent, such as the dignity of labor
and taking pride in ones work, we argue that the net effect of effort is still negative utility.
Otherwise, the agent would not need to be paid for his labor.
12. The difference between d
2
and d
1
narrows as e increases and the two are eventually
equivalent (i.e.,
2
d d
1
= ) above some critical level of e (See Appendix).
13. George A. Akerlof (1983), in one of the few papers by an economist on the subject of
changing type, argued that the moral education of children heavily impacted their
productivity in life.
34

14. In particular, Andrisani and Parnes (1983) assert that a goal of public policy should be to
assure that workers of all ages are able to expect payoffs to individual effort, hard work, and
investments in skills over the course of the life cycle.
15. The importance of moral and ethical training to the overall welfare of society has been
demonstrated in case studies of primitive cultures (Edward C. Banfield 1958) and modern
day democracies (Robert D. Putnam 2002). This research suggests that ethical norms form a
type of social capital that benefits the economic and social well being of a society. This
research also suggests that the variability in ethical sensitivity may be as large between
groups as between individuals within the same group.
35



FIGURE 1
Timeline of Events in the Model



Principal
specifies a
standard for
effort, d, and
wage,
Agent chooses to
accept or reject the
contract
Agent privately
chooses effort, a
Principal observes
performance
measure, X
Principal pays the
agent the wage,
Principal
observes
outcome, Y




36



FIGURE 2
Graph of First-Best Solution



pa
2
2
a
p a
FB
=
a
of Benefit
Effort:
of Cost
Effort:



















37



FIGURE 3
Graph of Flat Salary Solutions













( )
2
2 1
1 exp
log
rd
e
r
(

(
(







d
1
d
of Cost
Effort:
2
2
d
Salary Premium:
pa
of Benefit
Effort:



38



FIGURE 4
Graph of Incentive Solution



a
2
2
a
Risk Premium:
2
2
2
a
r pa
Effort:
of Benefit
Effort:
of Cost





















39



FIGURE 5
Comparison of the Cost of Incentive and Salary Solutions



a
Risk Premium:
( )
2
2 1
1 exp
log
rd
e
r
(

(
(

2
2
2
a
r
Salary Premium:
Region III Region II



















Region I




40

APPENDIX

Proof of Proposition 1

The flat salary, =
d
2
2
, will induce the standard level of effort from the agent (i.e., a = d) if it
satisfies the IR and IC constraints from equation 5. The flat salary satisfies the IR constraint
because,
( )
{ }
2 2
2 2
E 1 exp 0
d d
r
(
=

.
The flat salary satisfies the IC constraint when,
( )
{ }
2 2
2 2
0 maximum E 1 exp
a d
d a
r e


( )
2
2
0 1 exp
rd
e
1
2 1
1
log
r
e
d d

=
If , however, the IC constraint will not be satisfied and
1
d d >
( )
{ }
2 2
2 2
Argmax E 1 exp
a
d a
a r
(

e
0 a =

Proof of Proposition 2

If =
d
2
2
satisfies the IR constraint, then >
d
2
2
clearly does as well. The IC constraint is
satisfied when
41

( )
{ }
( )
{ }
2 2
2 2
E 1 exp maximum E 1 exp
a d
d a
r r

e
(


( )
( )
2
2
1 exp 1 exp
d
r r
(


e
( )
( )
2
2
exp exp
d
e r r
(



( ) ( )
2 2
2 2
exp 1 exp
d r
e r
( d
(

(


( )
( )
2
2 2
2
1 exp
exp
rd
d
r
e

(


(



( )
2
2
2 1
2
1 exp
log
rd
d
e
r

(

(

(

+

Proof of Proposition 3

According to Proposition 1, the principal can induce a level of effort without any salary premium
as long as the effort is below d
1
. We show in the body of the paper that the first-best effort is
d p
FB
= . Thus, for relatively low productivity firms, the first-best effort could be below the
agents reasonable level, d p d
FB
= <
1
. Thus, the optimal salary contract for these firms is the
first-best contract that pays the cost of the first-best effort.


42

Proof of Propositions 4 & 5

According to Proposition 3, when
1
p d the principal can induce the first-best effort d p
FB
=
without any salary premium, but not so when
1
p d > . According to Proposition 2, the cost of
inducing effort is:
1
d d
2
2
2
1
2
1 exp
log
rd
e
d
r

| |
|
\ .

(
( =
(

+
2 2
2
2
1
1 1
2
2
exp
log exp log
rd
e
d
rd
r r
| |
|
\ .

(
( | |
( = +
( |
( \ .

+
2
2
1
1
exp
log
rd
e
r
| |
|
\ .

(
( =
(


The principals problem then is to
1
Max:
d d

2
2
1
1
exp
log
rd
e
pd
r

| |
|
\ .

(
( =
(


The above objective function is concave in when
1
d d
( ) ( )
2
2 2
2
2
2 2
2
1 1
1 1
exp exp
0
rd rd d
rd

( (
+
( (
=
( (

>
( )
2
2
2
1 exp
rd
rd + <
Note that the left hand side of the above inequality increases linearly in d
2
while the right hand
side increases exponentially in d
2
. As such, this inequality holds for large values of d. The
critical question is whether it holds for
1
d d = because
( )
2
2
2
exp
rd
rd + < 1 if
1
d d
43

( )
2
1
2
1
2
1 exp
rd
rd +
( ) ( )
1 1
1 1
1 2 log
e e
+ .
Note that the left hand side of the above inequality increases in a concave manner in
( )
1
1 e

while the right hand side increases in a linear manner. As such, this inequality holds for e .
In summary, when e the wage function is convex in the domain , and when e
c
e
c
e
c
e
1
d d <
the wage function is initially concave and then becomes convex.
Case 1: e . In this case, the wage function is concave in and
c
e
1
d d
( )
2
2
1 exp
rd d
d
p

=

.
Therefore, the solution is the corner point d
1
when
1
1
0
d d d
d
p
e

= <
1
1
d
e
d p < .
In contrast, when
1
d
e
p > the solution is an interior point characterized by
( )
2 *
*
2
1 exp
rd
d
p

= .
Case 2: e . In this case, the wage function is initially concave and then convex in d.
c
e <

( )
d
2
d
2
1
2
d
1
d d
( )
2
2
2
2
1 exp
rd
d

1
d
e



44

Note that in this case the cut-off required for the interior point is lower than
1
d
e
. Specifically, the
solution is the corner point d
1
when
( )
2
2
2
1
1
2
1 exp
rd
d d
d p
e
< <


and an interior point (d
*
) characterized by
( )
2 *
*
2
1 exp
rd
d
p

= when
( )
2
2
2
2
1 exp
rd
d
p

> .
Note, however, that when
( )
2
2
2
1
2
1 exp
rd
p
d d
e
< <

there will be two interior values (a
minimum and a maximum) that will solve
( )
2 *
*
2
1 exp
rd
d
p

= , and the solution is the higher of
the two values.

Proof of Propositions 6

According to Proposition 3, the first-best contract is implementable if
2 1
1
log
r
e
p

. This is
equivalent to
(
2
2
1 exp
rp
e . If e = 0, then d
1
= 0 and the agent will never exert the
standard effort if he is paid only the cost of effort. Yet, if e = 0, then the salary premium goes to
infinity, implying that this solution will also not work. On the other hand, if d
1
> 0 or the salary
premium is finite, then it must be that e 0.
)
45

The Traditional Agency Problem

With ethical sensitivity constrained to zero in our model, the principals problem in
equation 5 becomes the traditional agency problem:

Max:

( ) E pa
Subject to:
IR
( )
{ }
2
2
E 1 exp 0
a
r
(



IC
( )
{ }
2
2
Arg max E 1 exp
a
a
a r
(


LC A BX = +
The LC constraint represents the traditional LEN assumption that the incentive contract is linear
in the performance measure. In the LEN framework the agents expected utility simplifies to
( )
2 2
2
2 2
1 exp A B B
a r
r a
(
+

.
The IC constraint thus simplifies to B a = . Substituting for B in the IR constraint results in
( )
2
2 2
2 2
1 exp A B 0
a a
r a r
(
+


2
2 2
2 2
A B
a a
a r + .
Note that A does not affect the induced effort. As such, the principal will chose A so that there is
no slack in the IR constraint. Therefore,
( )
2
2 2
2 2
A B E
a a
a r + = = .
46

You might also like