Essays on behavioral and organizational economics an experimental approach 2

showed that without delegating formal authority to the agent, the principal exerted higher than the Nash predicted effort levels while the mean effort exerted by the agent was not signif

CHAPTER I:

Introduction

Most economic models assume that economic agents are purely rational and self-interest In this view, people are motivated exclusively by own material self-interests The view can be traced back to Adam Smith In his “An Inquiry into the Nature and Cause of the Wealth of Nations”, Adam Smith argues that by rationally pursuing own economic interests in a free market environment, people can exhaust every mutually beneficial opportunities to produce goods and exchange with one another Such exhaustion of mutually beneficial exchanges thus results in full employment In many economic domains, the pure rational self-interest hypothesis has been quite successful in providing satisfactory explanations and accurate predictions to economic phenomena

Recently, countless evidence from many economic experiments has shown that the rational self-interest hypothesis could be too narrow to capture behaviors in many economic settings The results from these experiments have systematically refuted their corresponding theoretical predictions which are derived from the rational self-interest assumption This evidence has led many people to conclude that, in addition to maximizing their material payoffs, people also care about fairness and other people’s wellbeing, and exhibit reciprocity towards others This view is now reasonably well accepted in the discourse of economics Recent research has no longer focused on establishing whether people do exhibit other-regarding behavior, but instead on explaining the nature of this behavior Rabin (1993), Fehr and Schmidt (1999), and Bolton and Ockenfels (2000) are examples of such studies Problems that have puzzled economists for a long time, such as the persistence of non-competitive wage premium, have been explained by some of these theories

In this thesis, I investigate some issues of other-regarding preference using experimental methods Among many advantages that experimental method offers, it allows researchers to exert tight control over the environment under which decisions are made Recently, the use of experimental

methods has substantially enhanced our understanding of human behavior Many new insights that were not predicted by the theories have been uncovered Many new theories have been written to explain the observed results which in turn have shed lights on many observed economic and social phenomena

This thesis consists of four essays The first essay investigates the impact of other-regarding preference on behaviors inside organization based on the principal-agent relationship Its impact on behaviors in the society at large is examined in the second essay Specifically, the second essay investigates the characteristics and the nature of social norms in a social encounter where the unaffected third party observers are present The third and the fourth essays investigate the impact of institutional arrangement on the stability of other-regarding preference In what follows, I will briefly elaborate the contents of all those essays

In the first essay, the experimental quest is on the impact of other-regarding preference on behaviors inside organization More specifically, I study the role of delegation in the principal-agent relationship based on the theoretical construction of Aghion and Tirole (1997) The literature in the principal-agent theory mainly focuses on the use of monetary incentive in alleviating the agency problems Aghion and Tirole (1997), instead, emphasize the importance of using non-monetary incentives to curb managerial opportunism They argue that delegation of authority to an agent can be

a powerful mechanism to elicit high effort from the agent That is, when the agent is given the power

to decide which project to implement, he will work hard to ensure his preferred project is chosen For the principal1, the cost resulting from the delegation is that she loses some degree of control over the firm Despite this consequence, the benefits of allowing the agent to implement his preferred project can potentially outweigh the costs to the principal should a poor project be implemented without her being informed

I designed a simple experiment that closely follows the theoretical constructions of Aghion and Tirole (1997) I first parameterized their model and derived its theoretical predictions Using these parameterized values, I conducted several treatments of experiments The experimental results

1 From now on, referred to as she

showed that without delegating formal authority to the agent, the principal exerted higher than the Nash predicted effort levels while the mean effort exerted by the agent was not significantly different from the Nash prediction When formal authority was delegated to the agent, it resulted in a significant increase in the agent’s effort and a significant decrease in the principal’s effort However, delegation of formal authority resulted in higher payoff to the principal only when the principal’s interests were closely aligned with those of the agent’s Some interesting findings that are not predicted by Aghion and Tirole (1997) were also found in our experiment First, an uninformed principal did not always delegate the real authority to an agent Delegation of real authority resulted in higher payoff to the principal Further, an informed agent did not always propose the project that maximizes his own private benefit As also shown by many experimental works, both the principal and the agent seemed to exhibit a noticeable degree of fairness even at the expense of private costs in acquiring information on project payoffs With payoff from recommending a project randomly is dominated by the payoff from recommending inaction, the uninformed agent did not always refrain from choosing such an inferior strategy Lastly, I conducted a simple comparative statics analysis on the impact of the interests congruence on behaviors The simple comparative statics analysis showed that increasing the principal’s interests congruence resulted in the principal behaving less aggressively

in overruling her agent’s non-profit-maximizing project recommendations It also resulted in the agent behaving more selfishly

The experimental results suggest several possible implications to the design of organization First, the agent will be motivated to work hard when formal authority is delegated to him However, delegation of formal authority to the agent does not directly translate to higher profits to owners of the organization In spite of the inherent fairness that the agent holds, the agent is mostly self-interested and will take actions that benefit him first rather than the organization Delegation of formal authority

to the agent pays only when the principal’s interests are closely aligned with those of the agent’s Second, given that the agent might take excessively risky decisions, the benefit from delegating real authority to the agent could be severely discounted To the uninformed principal, however, delegation

of real authority to the agent still pays than retaining the authority

In the second essay, I investigate the characteristics and the nature of social norms using a modified third party punishment game (Fehr and Fischbacher, 2004) Recently, the study of social norms has kindled many economists Plethora of recent experimental evidence has shown that the affected second parties often punish someone who violates social norms Such punishment occurs at the affected second parties’ own expenses, most importantly, with the consequence that does not bring material benefit to them While the reasons why the unaffected third parties will punish the violations

of social norms are still not clear, it is undisputable that the punishment by the affected second party alone is not sufficient for the social norms to be enforced To this end, it is crucial that the unaffected third parties must have motivation to punish when social norms are violated (Bendor and Swistak,

2001, Carpenter and Matthews, 2002, Fehr and Fischbacher, 2004)

In their third party punishment game experiment2, Fehr and Fischbacher (2004) showed that the third party observers, whose material payoffs were unaffected by the violation of egalitarian distribution norm, were willing to enforce the social norm at their own expenses However, their design incorporated only a single third party observer In reality, social norms enforcement typically requires collective efforts from a population of third-party observers who are not directly harmed by the norm violations While the behavior of the single third-party has exhibited strong motive to enforcing the social norms, the same motive might not sustain when there are multiple observers

To explore these issues, I extend Fehr and Fischbacher (2004) by first introducing three (3) third party observers into the TPP game Including more than one (1) third party observer in the TPP game introduces temptation to free-ride on other third-party observers’ incentives to punish To the best of our knowledge, no research has analyzed the impact of increasing the number of third party observers on the incentives to punish the unfair dictator in the TPP game context The results of including three third party observers in the game show that incentive to punish, measured by the amount spent by the sanctioning third party observers, shrank significantly when the number of observers increased to three; interestingly, decision to punish measured by the proportion of third party observers chose to punish was not affected

2 From now on, referred to as TPP game

I also investigate the impact of punishment on behaviors In another treatment with multiple third party observers, I weakened the impact of punishment with the aim to analyze the impact of punishment on players’ behaviors I found that weakening the impact of punishment affected both the amount spent to punish as well as the decision to punish To complete the experiment, I also conducted a simple repeated dictator game treatment with random interactions among participants By comparing behaviors among different treatments of which the number of observers increased gradually from no observers, to one observer, and to three observers, I found a strong negative relationship between the number of third party observer and the amount given by the dictators

The third essay further evaluates the property of other-regarding behavior using a modified dictator game The central issue in the experimental quest is the impact of peer comparison of ability among allocators on their other-regarding behavior towards their respective recipient The baseline treatment is a setting in which information on the relative ranking of ability is not disclosed to the allocators If other-regarding behavior is inherently stable, it cannot be manipulated by changing the social environment Consequently, the presence of a social comparison of ability should not influence people’s other-regarding behavior If we observe any significant difference in the allocation decisions between these two underlying environments, we may attribute this difference to the presence of the social comparison of ability More importantly, it would also suggest that people’s other-regarding behavior is endogenously shaped by the social (institutional) environment Interestingly, the experimental results show that disclosing information on the relative ranking of the allocators’ ability and the distribution of their scores results in their giving a lower amount One possible conjecture is that being aware of the competitive environment and knowing that people tend to compare themselves against others may reinforce their sense of entitlement towards their own “hard-earned” money endowment and thus result in giving less The experimental results also suggest that a meritocratic environment that is characterized by an explicit social comparison will result in a society that exhibits less generosity towards others Similar to List (2007), the results also provide evidence on the manipulability of other-regarding behavior But there is one significant difference between this essay and List (2007) In the latter, the manipulation is done by changing the allocators’ action set, while in

this essay the manipulation is done by changing their social environment (institution) On a more general level, the experimental results suggest that a society that is excessively competitive and accustomed to social comparison will exhibit less generosity towards others

To check the robustness of the experimental results I also conducted an additional experiment with a distinct set of participants using a charity game, wherein the allocators must decide how much

of their hard-earned money to donate to a global charitable organization, Médecins sans Frontières (Doctors without Borders) Thus, instead of giving to an unknown recipient, allocators donate money

to a widely known charitable organization whose activities span the globe The results are similar to those obtained from our modified dictator game experiment, whereby the amount of giving is significantly reduced when the allocators are socially aware of their relative ranking of ability among their peers Finally, in both experiments, I also found evidence that, when subjected to a social comparison of ability, female allocators reduced their contribution rate significantly more than male allocators

The existing theories of other regarding preference share one common thread That is, they focus solely on factors that are directly related to the allocation decisions such as the motives and consequences behind those allocation decisions, but not on other factors that are seemingly unrelated

to, but could possibly influence those allocation decisions such as the procedure with which people earn their initial money endowment Rabin (1993) points out to the importance of the direct reciprocity motive A person’s behavior towards another person is driven by the (expected) intention

of the other person towards the latter Fehr and Schmidt (1999) and also Bolton and Ockenfels (2000) argue that other regarding behavior is primarily driven by the final distribution of payoffs among people in the reference group These theories disregard the possible link between the procedure with which the money endowment is generated and distributed among individuals and the allocation decisions of these individuals The fourth essay further investigates the possibility of this link using a two-stage dictator game

In the first stage, participants who were randomly selected as allocators had to perform a task

of answering GRE-GMAT like questions to earn money Better performance on the task resulted in

higher earnings From the first stage, I obtain the distribution of the allocators’ scores (performance)

on the task I then provide the allocators with the information on their individual ranking of scores among all allocators and the general distribution of scores Prior to playing the second stage, the allocators’ initial money endowment was determined In the baseline meritocratic treatment, the allocators’ initial money endowment was set equal to their individual earnings from the task In another treatment, which is call the egalitarian treatment, the allocators’ initial endowment was set equal to the average earnings of all allocators from the task This implies that in the egalitarian treatment the high performers were taxed, while the low performers were given subsidy by the high performers In the second stage, the allocators had to decide how much of their money endowment to give to other participants who were randomly designated as the recipients

The experimental results show that the allocators were more generous under the egalitarian treatment than under the meritocratic treatment Thus, in essence, the egalitarian treatment gave rise to

a stronger other regarding behavior of the allocators Surprisingly, being taxed and “forced” to subsidize the low performers did not diminish the high performers’ generosity The contribution rate and the absolute contribution of the high performers in the egalitarian treatment were not statistically different from the contribution rate and the absolute contribution of the high performers in the meritocratic treatment

However, the low performers who were at the receiving end of a “kind act” of the high performers significantly increase their contribution rate and absolute contribution given to the third party recipients in the egalitarian treatment Essentially, the generously egalitarian environment inculcates pay it forward attitude The low performers may have “felt obliged” to reciprocate for what they have received from the high performers by being more generous towards the third party recipients Boulding (1973) and Moody (2008) coin this form of reciprocity as serial reciprocity

CHAPTER II

Delegation of Formal and Real Authority in the Lab

Abstract

We experimentally study delegation of formal and real authority in the

principal-agent relationship based on the theoretical constructions of Aghion and Tirole

(1997) Our results show that the principal exerted higher than the Nash predicted

effort when she held formal authority while agent did not Delegation of formal

authority to the agent significantly increased the agent’s effort Unlike the theory

predicts, an uninformed principal did not always delegate real authority to an agent;

an informed agent did not always propose the project that maximizes his own

private benefit Furthermore, both the principal and the agent seemed to exhibit a

noticeable degree of fairness even at the expense of private costs Lastly, even with

the risk of suffering large wealth loss from randomly recommending a project, the

uninformed agent did not always refrain from engaging in such risky behavior

3 This research benefited from the research grant support R-122-000-107-112 awarded by the Faculty of Arts and Social Sciences-National University of Singapore

2.1 Introduction

Modern organizations are characterized by separation of ownership and control (Berle and Means, 1932) Despite having strong interests to be in control of a firm’s detailed operations, the owner (principal) of the firm tends to refrain from doing so The complexities of the organization and the spans of its operation imply that hiring professional managers (agent) to run the firm is almost inevitable However, given that the agent’s effort is not contractible and directly monitoring the agent

is difficult, empowering the agent to run the firm will only cause the agent to shirk from his responsibilities since working gives rise to disutilities The shirking problem becomes even worse when the interests of the agent and the principal are incongruent Instead of promoting the principal’s interests, employing the agent to run the firm gives the agent chances to advance his own private benefits This essentially is the well-known agency problem

With the root of the agency problem lying in the incongruent preferences, co-aligning the interests of the agent with those of the principal is fundamental to alleviating the problem Through a proper design of a compensation scheme, the agent can be incentivized to work in line with the principal’s interest However, an incentive scheme based on effort fails to work well since effort is mostly unobservable and hence not contractible Instead, vast literature in the principal-agent theory advocates the use of an incentive design based on measurable variables, such as the firm’s output A compensation scheme based on measurable contributions co-aligns the two parties’ interests It deters the agent from shirking since such behavior is nonprofit-maximizing to the agent

An alternative solution to reducing the agency problem is to strengthen the information system Although it does not directly change the agent’s preferences, strengthening the information system keeps the principal informed about whether the agent is behaving in the interests of the principal Consequently, deceiving the principal becomes more difficult

An elegant and innovative way to alleviate the agency problem is proposed by Aghion and Tirole (1997) In contrast to most principal-agent frameworks, their paper emphasizes the importance

of using non-monetary incentives to curb managerial opportunism They argue that delegation of

authority to the agent can be a powerful mechanism to elicit high effort from the agent Specifically, when the agent is given the authority to decide on the project to be implemented, he will work hard to ensure his preferred project is chosen For the principal, however, the cost resulting from the delegation is that she loses some degree of control over the firm Despite this, the benefits of allowing the agent to implement his preferred project can potentially outweigh the costs to the principal if the principal is uninformed and she chooses to implement a poor project

Aghion and Tirole (1997) also distinguish formal authority from real authority Formal authority is the authority which is legally conferred and stated in the contract It gives its holder the right to make decisions that affect the organization without being overruled by anyone inside the organization Real authority, on the other hand, is the effective control over decisions However, formal authority need not necessarily confer real authority The two types of authority coincide only when the holder of formal authority is informed about the prospects of available projects to be developed by the organization In this case, formal authority will be exercised by the informed authority holder When the holder of formal authority is uniformed, however, he or she will delegate real authority to the informed subordinates In this case, there is a separation of formal and real authority Such separation is necessary to avoid the uninformed formal authority holder from making worse decisions

Formal authority can also be delegated in an informal way Baker, Gibbons and Murphy (1999) show that, in most organizations, delegation of formal authority often represents an informal promise given by the principal to the agent As an informal promise, it can be withdrawn by the principal Similar to Aghion and Tirole (1997), they argue that delegation of formal authority can incentivize the agent to exert effort Unlike Aghion and Tirole (1997), however, an informed principal might be tempted to renege on the delegation decision by rejecting a project that is not in her best interest This, however, might not happen if the interaction between the principal and the agent is repeated In such a setting, the benefit deriving from the reputation of delegation can outweigh the gain from reneging on the promise of delegation Nor will the agent relentlessly implement his preferred project when the principal is uninformed The fact that the delegation of formal authority represents an informal promise implies that withdrawing the formal authority from the agent is

possible When the principal is uninformed, the threat of withdrawing the formal authority thus disciplines the agent This threat will prevent the agent from indulging in the excessive pursuit of his self-interests

Dessein (2002) shows that delegating formal authority to the agent can be better than retaining formal authority when the uninformed principal cannot verify the claims of a better informed agent This is because when the principal retains formal authority, it induces the agent to communicate biased information to the principal in order to advance his self-interests The communication can be in the form of exaggerating the returns of the agent’s preferred project or undervaluing the principal’s preferred project Information given by the agent is therefore noisy The quality of decisions that are made on the basis of this noisy information is thus poor Delegation of formal authority to the agent makes this strategic communication unnecessary to the agent However, the cost of delegation is the principal’s loss of control The trade off of delegating formal authority to

an agent is thus between a loss of control under delegation and a loss of information under communication Delegation dominates communication when the divergence in preferences between the principal and the agent is not too large relative to the principal’s uncertainty about the environment; otherwise, communication is better than delegation

While the theory of Aghion and Tirole (1997) is elegant, there has been no empirical testing

of the theory thus far It could be due to the lack of suitable data which makes econometric verification of the theory difficult Even if the data was available, controlling the confounding factors presents a major obstacle to an empirical testing An economic experimental analysis could offer an alternative way to test the theory By incentivizing real subjects in a controlled laboratory environment which mimics the theoretical construction, a test on the theory could be conducted This paper presents such an experimental analysis Its main objective is to study the role of delegation in a principal-agent interaction within the theoretical framework of Aghion and Tirole (1997)

There have been several experimental papers on the principal-agent models, however almost all of them focused on the use of direct monetary incentives to elicit effort from the agent A d r u b

n g t i a d G ä h e (

0 2

d m o s r t t a b h v o s o b a n d f o t e l b w e e c o e t t e

t e r t a p e i t n Specifically, by allowing the principal to offer her agent a contract chosen from any feasible combination of positive and negative fixed wages and a return share between 0 to

100 percent of the firm’s output, they found that principals tended to offer incentive compatible return shares and demanded negative fixed wages in many cases In return for the contracts offered, the agents tended to choose optimum effort levels in many cases However, they also found that contracts offered by the principals were fairer than the contracts predicted by the standard economic arguments

In particular, receiving generous offers resulted in a higher likelihood that the agent would exert above optimal effort levels than receiving excessive offers They argue that reciprocity explained why the agents deviated from exerting the optimal effort Thus, fair sharing between the principal and the agent is a behaviorally relevant consideration for the participation constraint in the context of a principal-agent relationship

Hamman, Loewenstein and Weber (2008) conducted repeated dictator games in a agent framework In one treatment of their experiment, the principal must decide the amount to be allocated to the recipient In another treatment, the principal could hire the agent to make the allocation decision on her behalf Upon knowing the decision made by the agent, the principal decided whether to keep using the same agent or to change to another agent for the subsequent round decision Further, the agents were only rewarded for being selected Comparing the result from the two treatments, they showed that principals behaved significantly less generously toward the recipients when they used the agents to make the allocation decision on their behalf They hence concluded that, besides the usual efficiency gains which the principal-agent theory advocates, a principal might hire

principal-an agent to take self-interested or immoral actions that the principal herself would otherwise be reluctant to take

Fehr, G c t r a d

i c s e g r (

1 9 ) showed that a robust reciprocity effect occurred when mutual reciprocity was possible between the owners of the firm and the workers Firms tended to reward workers who exerted more than contracted effort levels and punished workers who shirked Anticipating the firm’s reciprocal behavior, workers would exert higher effort than the optimal effort

level Reciprocal responses thus increased the set of enforceable contracts and resulted in large efficiency gains

Supported by countless experimental results from the dictator game, the ultimatum game and the gift exchange games, Fehr and Falk conclude that

“Contract theory and principal – agent theory, for example, typically restrict their attention to the motives to achieve income through effort and to avoid risks …… this narrow view of human motivation may severely limit progress in understanding incentives.” (Fehr and Falk (2002) page 687 – 688)

While the existing experiments have contributed to a better understanding of the agent theory, they examined mainly the role of direct monetary incentive in aligning the interests of principal and agent Experiments that addressed the role of non-monetary incentives are few Falk and Kosfeld (2006) studied the interaction between control and motivation in a principal-agent relationship In their experiment, a principal could control her agent by dictating a minimum effort level that the agent must exert Alternatively, the principal could choose to trust her agent by leaving the choice of effort completely to the agent’s discretion A comparison of the results from these two contracts showed that the agent was less motivated to perform when he was controlled than when he was trusted Controlling the agent enabled the principal to rule out any opportunistic behaviour, however, it significantly reduced the agent’s incentive to act in the principal’s interests This lack of incentive to promote the principal’s interests was caused by the agent’s perception of distrust when the agent was under controlled by the principal

principal-The central thesis of Aghion and Tirole (1997) is that delegation of formal authority incentivizes the agent to exert high effort Theoretical literature on the role of delegation is rich; however, experimental literature on delegation is few Fershtman and Gneezy (2001) showed that by delegating the decision right to a third party using an observable compensation scheme in an ultimatum game, a proposer was able to reach a significantly higher payoff than when the offer was

made by the proposer himself This was because the use of the agent as a delegate induced a change in the behavior of the responder which made the responder more inclined to accept the offer On the other hand, when the responder was allowed to delegate the decision on whether or not to accept the proposer’s offer, delegation became a commitment device for the responder to bargaining aggressively Delegation by the responder thus resulted in the proposer making a more generous offer

Another recent experimental study on the role of delegation was done by Bartling and Fischbacher (2008) Using a modified dictator game, they studied the moral responsibility attribution following delegation decisions In a one-shot encounter, the dictator could either decide on a fair division or an unfair division of income between the dictator and the recipient Alternatively, the dictator could delegate this decision to a second player whose payoff was perfectly aligned with the dictator Upon knowing the division of income allocated by the dictator or the delegate, the recipient, who was adversely affected by an unfair division, could decide to punish either the dictator or the delegated second party Their experimental results showed that when the dictator delegated the decision right to the delegated second party, the latter was harshly punished In contrast, the dictator was almost spared when the unfair division was allocated By delegating the decision right to the second party, the dictator thus shifted the blame to the delegated Knowing this, the dictator delegated more since it was profit maximizing to give the recipient an unfair division of income without being punished Their experiment showed that responsibility shirking or blame shifting was a powerful motive for the delegation of a decision right

While these experiments have offered fresh insights into the roles of delegation in decision making, they differ remarkably from the role of delegation in Aghion and Tirole (1997) In Fershtman and Gneezy (2001), delegation acts as an instrument to induce different belief and a commitment device to make bargaining aggressive In Bartling and Fischbacher (2008), delegation is a strategy to shift blame away from the delegator to the delegated second party In Aghion and Tirole (1997), however, delegation is used as an incentive device to induce effort Furthermore, the interactions in both Fershtman and Gneezy (2001) and Bartling and Fischbacher (2008) are not based on a principal-agent relationship Instead, they are based on modified ultimatum and dictator games To the best of

our knowledge, no experiment has been conducted to analyze the role of delegation as an incentive mechanism in the principal-agent interaction

We designed a simple experiment that closely follows the theoretical construction of Aghion and Tirole (1997) We first parameterized their model and derived its theoretical predictions Using these parameterized values, we conducted several treatments of experiments Our experimental results showed that without delegating formal authority to the agent, the principal exerted higher than the Nash predicted effort levels while the mean effort exerted by the agent is not significantly different from the Nash prediction When formal authority was delegated to the agent, it resulted in a significant increase in the agent’s effort and a significant decrease in the principal’s effort Some interesting findings that are not predicted by Aghion and Tirole (1997) were also found in our experiment First, an uninformed principal did not always delegate the real authority to an agent Further, an informed agent did not always propose the project that maximizes his own private benefit

As also shown by many experimental works, both the principal and the agent seemed to exhibit a noticeable degree of fairness even at the expense of private costs in acquiring information on project payoffs Lastly, with payoff from recommending a project randomly is dominated by the payoff from recommending inaction, the uninformed agent did not always refrain from choosing such an inferior strategy

Using the parameterized values, we also conducted some simple comparative statics analysis Our results from the simple comparative statics analysis partially confirms to Aghion and Tirole (1997) We found that without delegation, increasing the agent’s interest congruence has no impact on both the principal’s and the agent’s effort choice However, unlike the theoretical predictions, we found that the principal’s effort choice seemed to be unaffected by the change in her degree of interest congruence We also found that increasing the principal’s interest congruence resulted in the principal behaving less aggressively in overruling her agent’s non-profit-maximizing project recommendations

It also resulted in the agent behaving more selfishly

The paper is organized as follows Section 2.2 revisits the main theoretical constructions of Aghion and Tirole (1997) as well as their theoretical predictions Section 2.3 discusses the experimental design and procedures Experimental results and explanations are presented in Section 2.4 Section 2.5 concludes

2.2 The Aghion and Tirole (2007) Model and the Testable Hypotheses

In this section, we first outline the key theoretical constructions of Aghion and Tirole (1997) and the corresponding theoretical predictions Using these theoretical predictions, we then set up our experimental hypotheses These experimental hypotheses will then be evaluated in section 2.4

2.2.1 The Model

In Aghion and Tirole (1997), the organization consists of one principal and one agent In each period, the organization needs to decide whether or not to implement a project from n available projects choices, and n ≥ 3 The principal hires the agent to collect the information about the projects and implement the chosen project Each project is associated with a verifiable monetary payoffs of Bk

and bk; Bk for the principal and bk the agent The payoff configurations of these n projects are unknown Ex ante, both parties know that the maximum payoffs from the available projects are Band

b, B to the principal and b to the agent Out of these n available projects, there is one project which is the most preferred by the principal This project yields a payoff of B to the principal and private benefit of βb to the agent There is another project which is the most preferred by the agent This project yields payoff of α B to the principal and bof private benefit to the agent The payoffs

of these two projects are also known to both parties ex ante α β , ∈ (0,1 ] are degrees of interest congruence As αcloses to1, the closer the principal’s interests are aligned with the agent’s interests

As βcloses to1, the closer the agent’s interests are aligned with the principal’s interests Parties’ preferences are thus incongruent as long as αand βdeviate from 1

In order to identify the available projects’ payoff information, parties need to exert efforts,

E by the principal and eby the agent Effort is costly and both the principal’s effort cost functionCP( ) E and the agent’s effort cost function CA( ) e are strictly increasing and convex, whereE e ∈ , [ ] 0,1

The organization can be of two forms The form of organization dictates the ways by which the principal interacts with the agent In the P-form organization, the principal holds formal authority

In the A-form organization, formal authority is delegated to the agent The agent thus has the control over the project selection and implementation in the A-form The timing of interaction in the P-form organization is as follows,

P and A exert effort

simultaneously

Individually, they know

whether or not they are

informed

A proposes a project to implement to P

P can accept or reject A’s proposal If P accepts, A’s proposed project is implemented If P rejects (overrules A’s proposal), P asks A to implement a different project or to do nothing, which is denoted as project 0 and gives payoff of

0 for both

Outcomes are realized

Note: P: Principal, A: Agent

Figure 2.1: Timing of Interaction in the P-form organization

Immediately after exerting effort at t = 1, parties will be informed about the project payoff configurations With probability of E, the principal is informed about the project payoff configuration The agent is informed about the project payoff configuration with probability of e With these

probabilities of being informed, four possible states can be realized The associated probabilities of these states are shown in Table 2.1,

P informed P uninformed

A uninformed ( 1 e E − ) ( 1 − e )( 1 − E )Table 2.1: Possible Outcomes and their Related Probabilities

Both parties’ objectives are to maximize their individual payoffs When the agent is informed, with probability of e, he will rationally recommend his preferred project to the principal As one project among the n available projects pays sufficiently large negative payoffs to both parties, recommending a project randomly results in negative expected payoff The rational choice would be recommending inaction when the agent is uninformed Thus, with probability ( 1 e − ) that the agent is uninformed, the agent will recommend his principal inaction With probability of E when the principal is informed, she will exercise formal authority Any project recommendation that is not the principal’s preferred project will be overruled; instead, the principal’s preferred project will be imposed With complementary probability of ( 1 E − )when the principal is uniformed, she will rubber-stamp any project recommended by her agent so as to avoid picking a worse project Thus, while the principal holds formal authority, the agent exercises real authority when the principal is uninformed This is the delegation of real authority

Specifically, in the P-form, the principal maximizes her objective of

[1−αe B] =CP'( )E ( )3

and the best reply for the agent of

[1−E b] =CA'( )e ( )4

where CP'( )E and CA'( )e are the principal’s and the agent’s marginal cost of effort respectively

( )3 and ( )4 together are the basic trade-off of ‘loss of control and initiative’ That is, the principal supervises more, the higher her stake and the lower her level of interest congruence The agent works harder, the higher the monetary gain from his preferred project and the lower the principal’s interference

Formal authority can be delegated as a way to induce the agent exerting higher effort When the formal authority is delegated to the agent, the organization is called the A-form organization The timing of interactions in the A-form is described as follows in Figure 2.2,

P proposes a project to implement to A

A can accept or reject If A accepts, P’s proposed project is implemented If A rejects (A overrules P’s proposal), A insists on implementing his preferred project or to do nothing which is denoted as project

0 which gives payoff of 0 for both

Outcomes are realized

Note: P: Principal, A: Agent

Figure 2.2: Timing of Interaction in the A-form organization

In the A-form, the principal’s and the agent’s objectives are

[1 ] ( )

2.2.2 Hypothesis

Based on the theoretical predictions outlined in section 2.2.1, the following hypotheses are raised The first hypothesis aims to test whether the observed mean effort in the P-form converges to the theoretical predictions or not

H1: Both the principal’s and the agent’s effort are not significantly different from the Nash effort in the P-form organization

The second hypothesis aims to test the role of delegation of formal authority as an incentive mechanism to elicit high effort When the principal delegates formal authority to the agent, the objectives of the parties change from (1) and (2) to (5) and (6) As the agent now has the ultimate control over the project implementation, exerting high effort will have a direct consequence on whether his preferred project will be implemented Hence, a self-interested agent is motivated to work hard to ensure his preferred project is implemented Moving from the P-form to the A-form, the principal will have lower incentive to work hard This is because the principal anticipates that a self-

interested agent will turn down her project recommendation that maximizes the principal’s payoff Hence, following the delegation of formal authority, the agent’s effort choice will jump up while the principal’s effort choice will drop The effort responses following the delegation of formal authority form our next hypothesis

H2: Effort exerted by the principal will increase significantly and effort exerted by the agent will decrease significantly when the organization is shifted from the P-form to the A-form

Besides the effort choices, the theory also gives some interesting behavioral predictions These qualitative behavioral issues will form our subsequent hypotheses First, the informed principal will exercise her formal authority by overruling any project recommendations that do not maximize her payoff Following the overruling, the informed principal will impose her preferred project to be implemented since doing so is profit maximizing However, when the principal is uninformed, she will delegate real authority to her agent by rubber-stamping any project recommendation from her agent since imposing a project randomly results in negative expected payoffs The informed principal’s overruling decision and the uninformed principal’s decision to delegate real authority form the next three hypotheses

H3: An informed principal will overrule project recommendations that do not maximize her payoff

H4: Following the overruling, an informed principal will impose the project that maximizes her payoff

H5: An uninformed principal will accept her agent’s project recommendation

A self-interested agent, on the other hand, will always recommend his preferred project to the principal when he is informed When he is uninformed, however, he will recommend inaction since randomly recommending a project could result in negative expected payoff and is dominated by recommending inaction These two predictions form the last two hypotheses

H6: An informed agent will recommend his principal the project that maximizes the agent’s payoff

H7: An uninformed agent will recommend inaction to his principal

2.3 Experimental Design and Procedures

In this section, we first parameterize the model We assign values on the effort cost functions and the degrees of interest congruence Using these parameterized values, we derived the corresponding Nash effort levels We then outline the experimental design which mimics the theoretical constructions of Aghion and Tirole (1997) Finally, the experimental procedures which were implemented in our experiment are discussed

2.3.1 Parameterizing the Model and the Nash Equilibrium Effort

In the experiment, the principal’s effort cost function was specified as,

Based on the effort cost functions of ( )9 and ( )10 , the principal’s and the agent’s objective functions in the P-form organization are shown in ( )11 and ( )12

and

[1 ]

As in the P-form, the principal’s effort and the agent’s effort in the A-form are strategic substitutes in equilibrium Their individual effort choice is positively related to the payoffs of their preferred project, Bandb Further, the principal’s congruence factor has no impact on both parties’ effort choices; however, the agent’s effort choice is negatively related to his congruence factor

Using (13), (14), (17) and (18), the parameters required to compute the Nash effort are B, b,

αandβ They are the maximum payoff to the principal from her preferred project, the maximum payoff to the agent from his preferred project, and the principal’s and the agent’s interest congruence factors

We designed three experimental treatments Between treatments, we did not vary the principal’s best project payoff and the agent’s best project payoff The value of Band bwere fixed at 1.3 and 0.7 respectively for all treatments Thus, the variables that separate one treatment from the other are the degrees of interest congruence, αandβ In treatment 1, we haveα β= =0.2 Using treatment 1 as the baseline treatment, treatment 2 increases αto 0.8 and keeps β unchanged at 0.2 Treatment 3 increases β to 0.8 while α at 0.2

Substituting these parameterized values into ( )13 ,( )14 ,( )17 and ( )18 , the Nash effort levels can be easily computed Table 2.2 lists the parameterized values of α β, , ,B band their corresponding Nash effort in both the P-form and the A-form organizations

2.3.2 Experimental Design

This section details the experimental designs for project payoffs, the effort costs, the experimental Nash predictions and the timing of interactions in the experiment

a) Project payoff

In the experiment, there were six available projects to be chosen in each period They were Project A,

B, C, D, E and Project F Of these six projects, one project paid the maximum payoff to the principal; the other project paid the maximum payoff to the agent There was another project which paid sufficiently large negative payoffs to both the principal and the agent The presence of this project makes the expected payoff from implementing a project randomly negative The fourth project paid the fair returns to parties The fair project can only be chosen when the party is informed According the theory, the informed party will choose the project that maximize individual payoff The fair

project gives an average payoff of the two best projects They are respectively 1(1 )

2 +α B to the

principal and 1(1 )

2 +β b to the agent The last two of the available six projects conveyed the intention of punishment One project paid nothing to the agent and 70% of the principal’s maximum payoff to the principal The payoff structure allowed the principal to punish the agent at a cost of 30%

of her maximum payoff The other ‘punishment’ project paid nothing to principal and 70% of the agent’s maximum payoff to the agent The direction of punishment was therefore opposite In our experiment, the maximum payoffs from the available projects were fixed throughout the experiment

at B = 1.3and b = 0.7 For instance, if α = 0.2andβ =0.8, the project payoff configuration is shown in Table 2.3

Project C is the principal’s preferred project and Project F is the agent’s preferred project Project E pays sufficient negative payoffs to both participants The ‘fair’ project is Project A Project

B allows the principal to punish the agent and the agent can punish the principal by recommending or implementing Project D

As participants were paid in experimental points with conversion rate of one point equals to 2 Singapore Cents, using the project payoffs shown in Table 2.3 as experimental points makes the reward unattractive to incentivize participants To make the reward sufficiently salient, the theoretical payoffs were scaled by multiple of 100 Hence for each round of interaction, the principal was rewarded 1.3 Singapore dollars for implementing her best project and the agent was rewarded 0.7 Singapore if his best project was implemented A scaled payoff table shown to the informed participants is shown as follows in Table 2.4

In theory, the levels of effort represent both the intensities of effort exerted and the corresponding probabilities that players will be informed of the project payoffs Further, effort is a continuous variable of [ ]0,1 Using a continuous variable, however, is not practicable in the

experiment We therefore discretized the effort variable into 11 levels The distance between two successive effort levels was the same The objective of separating effort choices into 11 levels was to make effort choice decision straightforward to the participants Substituting these effort levels into the effort cost functions of ( )9 and ( )10 , the resulting effort cost tables are shown in Table 2.5 and Table 2.6

Table 2.6: Theoretical Effort Cost table for the Agent

Similar to the project payoffs, the theoretical effort costs involve decimal numbers With the conversion rate of 1 experimental point equals to 2 Singapore Cents, directly using these effort costs makes effort trivially inexpensive to the participants In line with the treatment of scaling the project payoffs, we scaled the effort costs by multiple of 100 In this way, exerting cost is no longer inexpensive

Substituting the scaled effort costs back to the effort cost functions, the effort choices will be some discrete numbers between 0 and 10 It is important to note that both in the theory and in the experiment, levels of effort represent the intensities of work exerted by the participants Exerting 0 level of effort represents complete shirking Exerting 10 level of effort in the experiment, or 1 in the theory, means working at the highest possible intensity Whether e E ∈, [ ]0,1 or e E ∈, [0,10], the same message is conveyed to the participants without loss of generality

As stated in Aghion and Tirole (1997), each effort level is associated with a probability of being informed In our experiment, the same construction remained For instance, participant

choosing 0 level of effort would never be informed about the project payoff table; participant choosing 5 level of effort would have a probability of 0.5 to be informed about the project payoff table; participant choosing 10 level of effort would be informed with certainty Hence, a complete transformed set of effort cost tables are shown in Table 2.7 and Table 2.8 Table 2.7 and Table 2.8 are the effort cost tables we used in the experiment

Table 2.8: Effort Cost table for the Agent

c) Experimental Nash predictions

Using the project payoffs shown in Table 2.4, effort costs shown in Table 2.7 and Table 2.8, the corresponding Nash effort levels are shown in Table 2.9 The effort choices listed in Table 2.9 were used to test Hypothesis 1

Stage 1 Allocation of authority

At the beginning of each period, participants were informed about the authority structure of either the P-form or the A-form In the experiment, the P-form organization was named as Part-1 of the experiment and the A-form organization was named as Part-2 of the experiment This was designed to avoid framing effect

Participants were then informed that there were six projects available for each period and the payoffs of the two best projects Participants were also informed that one of these six projects paid extremely negative payoffs to both parties Further, participants would be asked to leave the experiment if their accumulated earnings were negative

Stage 2 Exerting effort to find out project payoff information

In each decision period, the principal was given 80 endowment points and the agent was given 50 endowment points The effort cost tables of Table 2.7 and Table 2.8 were shown to all participants so that participants were informed about the principal’s and the agent’s effort cost structures Using the endowment points, the principal and the agent simultaneously exerted their desired efforts to find out the project payoff information Depending on the level of effort chosen, effort costs based on the effort cost tables were deducted from the initial endowments

Stage 3 Project recommendation

The project payoff information table was revealed to participants based on the effort exerted and the corresponding probabilities shown in the effort cost tables

Under the P-form organization, whether the agent was informed about the project payoff information or not, the agent had to recommend his principal a project to be implemented The agent could also recommend the principal inaction which was denoted as Project 0 The payoffs to both

parties were 0 if Project 0 was implemented Under the A-form organization, the principal had to recommend the agent a project to be implemented The principal could also recommend Project 0

Stage 4 Project selection

Following the project recommendation, the party holding formal authority could choose either accepting the recommended project, implementing a different project or implementing Project 0

Stage 5 End of period

The chosen project was implemented and its payoffs were allocated to the respective parties

The above 5 stages completed one round of interaction At the end of each round, participants were informed about their payoffs from the existing round as well as their accumulated payoffs from previous round(s) The same game was repeated thereafter The experiment ended with the experimenter stopped the experiment At the end of experiment, the cumulative points from all rounds were converted to cash and were paid to participants immediately in private The conversion rate was

1 point to 2 Singapore Cents and this was made known to all participants before the experiment started

For each treatment, 12 repeated rounds of the P-form organization followed by 11 repeated rounds of the A-form organization were conducted Round 1, 2 and 13 were trial rounds Earnings from the trial rounds were not counted The payoffs of the six available projects were the same in all the decision rounds during the experiment However, the project identities were randomly rearranged

as the game repeated Switching the project identities randomly in each round was necessary to prevent participants from guessing the project identities without exerting true effort In this way, the equilibrium effort choices could be elicited from participants towards the last few rounds of the game

In our experiment, a ‘soft information’ setting was adopted In a ‘soft information’ setup, at stage 5, the principal would not be able to verify the recommended project’s payoffs when she was uninformed In the A-form organization, the agent would not be able to verify the recommended project’s payoffs when he was uninformed With a ‘hard information’ setup, the recommended project’s payoffs will be known to the party with formal authority regardless of whether he or she is informed or not We suspect that using the ‘hard information’ setup might induce the recommending party to randomly recommend a project without exerting effort since the project that pays negative payoffs will always be rejected

e) Experimental procedures

The experiment was conducted using the www web at the National University of Singapore Faculty of Arts and Social Science Computer Lab The computer system allowed simultaneous multiparty interactions via internet The information from the decisions made by participants was transferred between the system administrator and matched partners The transfer of this information was instant

Upon registration, participants were randomly allocated a computer Communication was strictly not allowed inside the computer labs Experimental instruction was read out loud before the experiment started Thereafter, a PowerPoint slideshow of the computer screenshots which described the decision interfaces was shown to all participants Following that, participants were required to answer a set of questionnaires designed to enhance and test their understanding of the experiment When all the questions were correctly answered, experimenter then distributed a randomly generated password which enabled participants to log into the experimental system Once participants logged in

to the system, they were informed about their roles in the experiment Two roles were randomly assigned by the computer system, they were Employer (the principal) and Employee (the agent) Once the role was assigned to the participant, it was fixed throughout the experiment However, participants’ interacting partners were matched randomly across rounds and all participants were informed about this information The random matching of partners across rounds gave participants

opportunity to learn from previous experience without building reputation As interaction between the principal and the agent was anonymous, the interaction thus mimicked the one-shot game

In each session, there were 23 decision rounds Across each round, we switched the identities

of the projects randomly The first two rounds and round 13 were practice rounds and they were not taken into account in the calculation of payoff Participants were not paid for the practice rounds To avoid the end game effect, participants were not told the total number of rounds to be played for each organizational form They kept playing the game until experimenter stopped them

We started with the P-form experiment for the first 12 rounds and followed by the A-form experiment for the rest rounds The comparison in participants’ decisions made between the P-form and the A-form constitutes a within subject comparison This allows us to assess the impact of delegation of formal authority on effort choices The experimental sequence to start from the P-form organization then to the A-form organization was not informed to the participants

At the end of each round, participants were informed about their own earning for the round as well as the accumulated earnings from previous round(s) The information about their partners’ effort choices and earnings were private information to them If negative earning occurred, the affected participant would be expelled from the experiment and the position was replaced by a standby At the end of experiment, their cumulative earnings were paid in private immediately

In total, we conducted 6 sessions of the experiment with the following arrangements,

Number of Participants Session Treatment

To avoid the impact of possible end game effect on the experimental analysis, we did not include data obtained from round 12 and round 23 for the experimental analysis in Section 2.4 Excluding round 1, 2 and round 12, we thus have 216 interactions between the principal and the agent for treatment 1, 216 interactions for treatment 2 and 207 interactions for treatment 3 in the P-form organization Excluding round 13 and round 23, we have 216, 216 and 207 interactions for treatment

1, 2 and 3 in the A-form organization

Participants were undergraduate students from the National University of Singapore with various academic backgrounds They were recruited from campus-wide advertisement No participant was allowed to participate more than one session of experiment On average, each participant earned

20 Singapore dollars including the 5 Singapore Dollar show-up fees in 1.5 hours

2.4 Experimental Results

This section presents the experimental results We first present the experimental results on the principal’s and the agent’s effort choices We then evaluate the impact of the delegation of formal authority on the principal’s and the agent’s effort choices Following that, we examine the unformed principal’s decision to delegate real authority and the informed principal’s decision to overrule Thereafter, we look at the informed agent’s decision to recommend his self-interested project and to recommend the fair project We finally analyze the uninformed agent’s tendency to engage in gambling behavior

Countless experiments have shown that behaviors differ as the game repeats The public good experiment (Ledyard, 1995), for instance, shows that players’ contribution levels decayed downward after the first round, but remained above zero toward the end of the game In our experiment, participants played the same game repeatedly over 13 rounds in the P-form organization and 12 rounds in the A-form organization with randomly matched partners The random matching of partners ruled out the possibility of reputation building among the players Repeating the game with sufficiently large number of rounds allowed the players to accumulate learning As a result, equilibrium behaviors could be elicited at the later rounds of the game Using the effort level elicited from round 11, Result 1 shows whether the experimental effort coincided to the hypothesized Nash equilibrium effort

Result 1 Both the principal’s and the agent’s effort choices are approaching the Nash predictions as the game repeats In equilibrium, however, the principal’s effort is still significantly higher than the Nash predicted effort while the agent’s effort is not statistically different from the Nash

Figure 2.3, Figure 2.4 and Figure 2.5 show the mean effort choices made by the principals and the agents across rounds from all the three treatments

Figure 2.3: Mean Effort Choices in Treatment 1

Figure 2.4: Mean Effort Choices in Treatment 2

Figure 2.5: Mean Effort Choices in Treatment 3

In all the three treatments, two common observations are shown by the above graphical presentations of effort choices First, the principal starts with an effort level that is close to the Nash predicted effort levels It then evolves further closer to the Nash predicted effort towards the latter rounds However, the agent’s effort choice decays considerably in the latter rounds Starting with a very high effort level, it gradually declines and evolves to the Nash predicted effort towards the latter rounds Such a pattern suggests that learning could be taking place among the players as the game progressed The learning might thus explain why the effort choices at the latter rounds of interaction are considerable closer to the Nash predicted efforts

Although the participants did not know that the P-form organization ended at round 12, as a prudent concern to avoid end game effect, we believe that the data elicited from round 12 should not

be used to perform the statistical analysis To test whether the players’ effort choices evolved to the Nash predicted effort, we take the effort choices made by players at round 11 as the sample effort choices and test them against the Nash predict effort using the Sign Test and the Wilcoxon Signed Rank Sum Test (two-sided) The Sign Test is used to test the null hypothesis that the mean of a distribution is equal to some value without assuming that the distribution is normal The test uses Binomial distribution to measure the probability of the observed values that are more than or less than

the hypothesized value The Wilcoxon Signed Rank Sum Test is a more powerful test Not only it takes into account of the sign of the difference between the observed values and the hypothesized value, it also considers the magnitude of the difference The results from these tests are reported in Table 2.11

The results from the Sign Test and the Wilcoxon Signed Rank Sum Test show that, at 5% significance level, the principals’ effort choices are all significantly higher than the predicted Nash effort levels However, the agents’ effort choices are not significantly different from the Nash effort levels This happens to all the three treatments of experiment we conducted Comparing the actual effort choices and the predicted effort choices from all the three treatments shows that the principal worked harder than the self-interested profit maximizing effort levels while agent did not Thus the observed behaviors only partially support H1

1 ** Significance at the 5% level

2 Significance at the 10% level

Table 2.11: P-values for the Sign test and the Signed Rank test (two-sided)

Why did the principal exert effort that is higher than the predicted effort level? We believe this could be due to the principal’s lack of trust to the agent In the experiment, the same principal was matched to a randomly chosen agent to form an organization in each round of interaction The random matching of partners made the game similar to the one-shot game even though the same game was

repeated with multiple rounds Further, the interaction between the principal and the agent was anonymous The anonymity of the players’ identities implies that the principal was not able to monitor the agent when the game was being played As a result, punishing or rewarding the agent become groundless In such a setting, it is difficult to build trust between parties

Without putting much trust on her agent, it became necessary for the principal to work harder

in order to avoid over-relying on the agent’s project recommendation; that is delegating real authority

to the agent In particular, when the expected payoff from randomly implementing a project was negative, delegating real authority to the agent could result in severe loss if the agent behaved riskily The only way for the principal to alleviate these problems was to work hard enough so as to be informed about the project payoffs

The same explanation could also be interpreted as ‘the principal’s fear of losing control’ Delegating real authority to the agent involves a loss of control over the project implementation If the principal did not work hard enough, she was less likely to be informed about the available projects’ payoffs and more likely to be forced to accept the agent’s project recommendation or to implement

‘Project 0’ In either case, the principal’s preferred project would not be implemented Delegation of real authority to the agent was a reluctant decision to be made The only possible way to avoid making such a decision was to work harder

In the P-form organization, the principal is the holder of formal authority When formal authority is delegated to the agent, the organization is switched to the A-from The main thesis of Aghion and Tirole (1997) is that delegation of formal authority to the agent increases the agent’s incentive to acquire the project payoff information at the expense of the principal’s losing control This is because delegation of formal authority to the agent changes the objectives of both parties With the agent holds formal authority, the amount of effort exerted by the agent will have a direct consequence to whether or not his preferred project will be implemented Hence, the agent is motivated to work harder However, the principal knows that a self-interested agent will always turn down her recommendation of projects that do not maximize the agent’s payoff, her incentive to

acquire the project payoff information is thus reduced Given that effort is costly, the principal will then work less hard Hence, when the principal delegates formal authority to the agent, we should expect the agent to exert higher effort and the principal to exert lower effort than in the P-form organization Our next result confirms this,

Result 2 Delegation of formal authority to the agent significantly increases the agent’s effort and significantly decreases the principal’s effort

As shown in Figure 2.3, Figure 2.4 and Figure 2.5 that, in all the three treatments, the agents’ effort choices jump up significantly when the organization is switched from the P-form to the A-form

In contrast, the principals’ effort choices drop significantly Further, comparing the effort choice made

at the latter rounds of the P-form organization against the effort choices made at the latter rounds in the A-form organization, both the increases in the agent’s effort choice and the decreases in the principal’s effort choices are stable changes of behaviors The stable changes in effort levels thus lend significant supports to H2

Table 2.12 details the payoffs to the principal in the P-form organization and in the A-form organization The payoffs in the P-form are the average of accumulated payoffs from round 3 to round

11 and the payoffs in the A-form are the average of accumulated payoffs from round 14 to round 22

Note:

*Significance at the 5% level

Table 2.12: Mean Payoffs Per Round

Overall, delegation of formal authority does not seem to be profitable to the principal The mean payoff to the principal in the P-form is 120.67 per round, whereas in the A-form, this is 102

Their difference is significant at the 5% level based on the Mann–Whitney–Wilcoxon Test Categorizing the payoff information according the treatment, however, one can find that treatment 3 resulted in a higher payoff to the principal in the A-form; whereas in treatment 1 and treatment 2, the payoffs to the principal are higher in the P-form than in the A-form The higher payoff in treatment 3 could be due to the principal’s high interest congruence As αcloses to1, the principal’s interests are more aligned with the agent’s interests When formal authority is delegated to the agent, the agent implements his self-interested project when he is informed In treatment 3, however, the agent’s preferred project still pays 80% of the principal’s preferred project payoff to the principal Whereas in treatment 1 and treatment 2, the same project only pays the principal 20% of her preferred project payoff Thus, delegation of formal authority can be financially attractive when the principal’s interests are closely aligned with the agent’s interests

Aghion and Tirole (1997) also distinguish formal authority from real authority Formal authority gives its holder the right to make decisions which affect the organization without being overruled by anybody inside the organization Real authority, on the other hand, is the effective control over decisions In the P-form organization, formal authority coincides to real authority when the principal is informed about the project payoff information When the principal is uniformed, however, the agent exercises real authority That is, the uninformed principal delegates real authority

to the agent with probability of one Our next result shows that this probability is less than one

Result 3 The uninformed principal does not always delegate real authority to the agent

When the uninformed principal delegates real authority to the agent, she accepts the agent’s project recommendation Of data collected from round 3 to 11 in the P-form, there are 140 observations of which the principal was uninformed Out of these 140 observations, delegation of real authority occurred 94 times Some principals delegated as long as she was uninformed, whereas some others never delegated To assess the mean propensity of the uninformed principal’s decision to