《Dessein_Authority_Communication_Password_Removed.docx》由会员分享，可在线阅读，更多相关《Dessein_Authority_Communication_Password_Removed.docx（29页珍藏版）》请在得力文库 - 分享文档赚钱的网站上搜索。
1、Review of Economic Studies (2002) 69, 811838 0034-6527/02/00320811$02.00 c 2002 The Review of Economic Studies Limited Authority and Communication in Organizations WOUTER DESSEIN University of Chicago and CEPR First version received October 1999; final version accepted January 2002 (Eds.) This paper
2、 studies delegation as an alternative to communication. We show that a principal prefers to delegate control to a better informed agent rather than to communicate with this agent as long as the incentive conflict is not too large relative to the principals uncertainty about the environment. We furth
3、er identify cases in which the principal optimally delegates control to an intermediary, and show that keeping a veto-right typically reduces the expected utility of the principal unless the incentive conflict is extreme. 1. INTRODUCTION This paper is concerned with the old saying that knowledge is
4、power. In organizations, much of the information used in decision making is dispersed in the hierarchy. Lower-level managers, for example, are often much better informed about consumer needs, competitive pressures, specialized technologies or market opportunities than their superiors. The financial
5、press is full of stories about how companies have pushed decision rights lower in the hierarchy in order to profit from this local knowledge.1 For the same reason, newly acquired subsidiaries are often left with substantial autonomy. The goal of this paper is to better understand why an uninformed p
6、rincipal (the company owners, senior management) may grant formal decision rights to an agent (senior or middle management) who is better informed but has different objectives. We argue that a principal often delegates authority in order to avoid the noisy communication, and hence the loss of inform
7、ation, which stems from these differences in objectives. At first sight, it may seem a puzzle why keeping authority and letting the agent report would not always weakly dominate delegation. By keeping authority, the principal has always the option to rubberstamp the proposals of the agent, but she m
8、ay also refrain him from implementing projects which are obviously not in the interest of the organization. By delegating authority, in contrast, the principal commits to never reverse the agents decisions. We will nevertheless argue that delegation is typically a better instrument to use the local
9、knowledge of the agent than communication. Key to our analysis is that differences in objectives between principal and agent are often systematic and predictable. It is, for example, well documented that managers may be short-term biased, status-quo biased, risk-averse, empire builders etc. Whenever
10、 the principal and the agent systematically disagree on a certain action dimension, the principal will not rubberstamp a naive recommendation by the agent of his preferred action, but try to correct for the bias in objectives. As the agent is not naive but anticipates this, communication is then inh
11、erently strategic andin equilibriumnoisy. Hence, the central trade-off in our paper is one between a loss of control under delegation and a loss of information under communication. 1. Among firms decentralizing decision rights in the 1990s are AT&T, General Electric, Eastman Kodak, Fiat, Motorola, U
12、nited Technologies, Xerox and, recently, Ford. 811 812 REVIEW OF ECONOMIC STUDIES Model. In order to analyse this trade-off, we develop a stylized model in which the principal (she) must screen among a range of projects which differ from each other on one dimension. The agent (he) has superior infor
13、mation on which project is best for the principal, but his objectives differ in a systematic way. He could, for example, always prefer a larger project than the principal (size-bias). For simplicity, this bias is constant and positive. Section 3 provides a discussion of the kind of biases we have in
14、 mind. The private information of the agent is assumed to be soft, that is the agent cannot prove or certify his knowledge. Furthermore, following Grossman and Hart (1986) and Hart and Moore (1990), we posit that projects (actions) cannot be contracted upon and, hence, the principal cannot use a sta
15、ndard mechanism to elicit the private information of the agent. The principal, however, can contract on the authority over the project. Indeed, to engage in a project, some critical resources are needed which are controlled by the principal. This implies that the agent normally needs the fiat of the
16、 principal to implement a project, but the principal can also delegate decision rights to the agent by granting him the authority over the use of these critical resources. In our organization, the principal thus faces the choice between fully delegating a task to a better informed agent or to order
17、the latter what to do after having consulted him. If she keeps decision rights and consults the agent, a game of strategic communication takes place in which each equilibrium is characterized by a partition of all possible states of nature and where the agent introduces noise into his signal by only
18、 specifying to which partition element the realized state of nature belongs. Given the information provided by the agent, the principal then takes the action which maximizes her expected utility. Such a strategic information transmission has been first analysed by Crawford and Sobel (1982), hereafte
19、r referred to as C S. While communication always involves a loss of information as long a preferences are not perfectly congruent, a central result of their paper is that the closer the preferences of agent and principal, the better is communication. The loss of information even goes asymptotically
20、to zero when differences in objectives disappear. Delegation, in contrast, results in a loss of control since the agent always takes a decision which is biased relative to the first best. Similarly, this loss of control becomes smaller when the agents preferences are closer to those of the principal
21、 and disappears in the limit. At first sight, the optimal allocation of authority is thus not trivial. Results. Our main finding is that the principal optimally delegates control as long as the divergence in preferences is not too large relative to the principals uncertainty about the environment. T
22、hus, if the state of nature is uniformly distributeda standard assumption in almost any application of the CrawfordSobel modelthe principal prefers delegation to communication whenever the agents bias is such that informative communication is feasible. The larger the uncertainty about the environmen
23、t, the larger is the range of biases for which the principal delegates control. More generally, for any given information structure, delegation dominates communication if the bias is sufficiently small. Indeed, for any continuous and twice differentiable distribution, as the agents bias tends to zer
24、o, a principal who keeps control and communicates, will take an action which is on average an infinite times further away from the first best than the action the agent would take. The intuition behind these results lies in the nature of the screening mechanism at work. For the agent to be induced to
25、 tell the truth, it is typically necessary that his messages become increasingly noisy as they recommend actions which go further in the direction of his bias. Intuitively, if an empire-builder recommends a large project, this message is less informative than if he recommends a small project. In add
26、ition, to prevent the agent from exaggerating his information, the increase in noise in subsequent signals should also be proportional to his bias. As a result, the better is communication, and thus the more messages the agent is able to send, the larger must be the average noise in these messages r
27、elative to the bias. In terms of the DESSEIN AUTHORITY AND COMMUNICATION IN ORGANIZATIONS 813 C S-equilibrium, the finer the equilibrium partition becomes in absolute terms, the more coarse it is relative to the bias. Hence, the more informative is communication, the better it is to delegate authori
28、ty to the agent and avoid communication. In contrast, if we keep the bias constant, but change the information structure, communication dominates delegation if the uncertainty about the environment is sufficiently small. Indeed, while changing the information structure does not affect the loss of co
29、ntrol under delegation, a more precise prior allows the principal to select an action which is on average much closer to the first best, and thus substantially reduces the loss of information under communication. While informative communication may then dominate delegation (e.g. when the prior of th
30、e principal is very steep and b is not too large relative to the support), communication is then typically very noisy. Simulations with (truncated) normal distributions and quadratic loss functions show that delegation is optimal unless the bias is so large that communication is almost uninformative
31、: regardless of the variance of the distribution, only if the bias is such that the agent recommends the same action in more than 98% of all states of nature, communication does better than delegation. To conclude, we investigate whether the principal can improve upon the pure delegation outcome by
32、some limited forms of delegation: In Section 6, we show that in the leading example of Crawford and Sobel,2 for moderate biases, the principal optimally delegates decision rights to an intermediary with objectives in between hers and her agents. Doing this, the principal may prevent the agent from i
33、mplementing extreme projects, without jeopardizing too much the communication concerning small and intermediate projects. Delegation to the agent remains optimal for small biases. In Section 7, finally, we consider delegation with veto-power for the principal, a mechanism which is known as the close
34、d rule in Political Science (see Section 2). Again, one might conjecture that delegating but retaining veto-power should at least weakly dominate complete delegation. For reasonable choices of the status quo and a uniform distribution, however, we show that keeping veto-power is only beneficial for
35、large divergences of preferences. For small or moderate biases, keeping veto-power results in additional variance in the deviation from the first-best, and complete delegation is optimal. Outline. The paper is organized as follows: Section 2 gives an overview of the related literature. Section 3 des
36、cribes the model. Section 4 characterizes the equilibrium for given decision rights. Section 5 then analyses the optimal allocation of authority. We go on to investigate the value of delegating control to an intermediary (Section 6) and the value of keeping veto-power (Section 7). Section 8, finally
37、, discusses various extensions of our model: profit- sharing arrangements, private information concerning the agents bias, repeated interaction, and verifiable information. We conclude in Section 9. 2. RELATED LITERATURE The incentive view on delegation. So far, the economic literature on organizati
38、ons has emphasized an incentive based rationale for delegation. In particular, Aghion and Tirole (1997), show that a principal may delegate formal authority to an agent in order to give the latter better 2. In the leading example of C S, principal and agent have a quadratic loss function and the sta
39、te of nature is assumed to be uniform on 0,1. So far, it has been the working horse for almost any application of C S. 814 REVIEW OF ECONOMIC STUDIES incentives to acquire information.3 While the focus of Aghion and Tirole is on the impact of authority on the information structure, we take the infor
40、mation structure as giventhe agent is assumed to be better informedand we investigate how the allocation of authority affects the use of this private information, providing a purely informational rationale for delegation. The informational theory of legislative rules. Both the incentive based ration
41、ale and our purely informational rationale for delegation have a counterpart in the informational theory of legislative rules in political science. In a very influential paper, Gilligan and Krehbiel (1987) adopt the leading example of C S4 in order to provide a rationale for the use of restrictive a
42、mendment procedures employed in the U.S. House of Representatives. In particular, Gilligan and Krehbiel (G K ) is concerned with the motivation of the closed rule, under which an uninformed legislature can only veto but not amend a proposal of a committee. If a proposal is vetoed, the status quo (wh
43、ich is taken exogenous) prevails. As Aghion and Tirole, G K assumes that this committee must make an effort to become informed. For the leading example of Crawford and Sobel, G K finds that when the preference divergence between the legislator and the committee is small, the closed rule is to be pre
44、ferred over the open rule under which the legislature can freely amend the proposal of the committee. While G K mainly emphasize the closed rules impact on incentives to acquire information,5 Krishna and Morgan (2000) have recently shown that even in the absence of an information acquisition problem
45、, the closed rule dominates the open rule as long as informative communication is possible under this open rule, a result which is similar to ours. While delegation is related to the closed rule,6 Section 7 shows that in the model of G K , for reasonable choices of the status quo, delegation strictl
46、y dominates the closed rule unless preferences divergences are extreme. In this sense, our paper points to a missing element in the reasoning of G K , which should explain why the committee then does not receive full decision power. Other related literatures. Finally, our paper differs from a number
47、 of literatures on information revelation in its assumptions on the commitment ability of the principal. By adopting an incomplete contracting approach, we clearly depart from the standard principal agent model in which the principal elicits private information by designing a mechanism. While this a
48、pproach may explain many institutions, we feel that the underlying premise that the principal can perfectly and without cost commit herself to any mechanism is too strong in many organizational contexts. In sharp contrast with the mechanism design literature, the literature on strategic communicatio
49、n or cheap talk, initiated by CS assumes that no commitment at all is possible. When we think about communication in organizations, this is also rather unrealistic. Indeed, it is an insight of the property rights literature that ownership or more generally the control over critical resources confers authority to its holder. Hence, the principal can commit to never overrule the decisions of the agent by delegating to him the control over these critical resources. We discuss some of the recent literature on cheap tal
黑公网安备:91230400333293403D