Read the article: Bolton, P., Brunnermeier, M. K., & Veldkamp, L. (2013). Leadership, Coordination, and Corporate Culture.Review Of Economic Studies,80(2), 512-537.

Based on the article findings, do you believe a leader coming into an existing organization, can change the culture of the organization? How?

The Review of Economic Studies Advance Access published December 22, 2012 Review of Economic Studies (2012) 0, 1-26 doi:10.1093/restud/rds041 The Author 2012. Published by Oxford University Press on behalf of The Review of Economic Studies Limited. Leadership, Coordination, and Corporate Culture PATRICK BOLTON Columbia University MARKUS K. BRUNNERMEIER Princeton University and LAURA VELDKAMP New York University First version received November 2009; final version accepted July 2012 (Eds.) What is the role of leaders in large organizations? We propose a model in which a leader helps to overcome a misalignment of followers' incentives that inhibits coordination, while adapting the organization to a changing environment. Good leadership requires vision and special personality traits such as conviction or resoluteness to enhance the credibility of mission statements and to effectively Downloaded from http://restud.oxfordjournals.org/ at Columbia University Libraries on April 9, 2013 rally agents around them. Resoluteness allows leaders to overcome a time-consistency problem that arises from the fact that leaders learn about the best course of action for the organization over time. However, resoluteness also inhibits bottom-up information flow from followers. The optimal level of resoluteness depends on followers' signal quality and the corporate culture of the organization. Key words: Leadership, Coordination, Overconfidence, Corporate Culture JEL Codes: D23, D7, D81 "One of the most important prerequisites for trust in a leader is steadiness. The need for reliability is not only ethically desirable, it is generally a practical necessity. A leader who is unpredictable poses a nerve-wracking problem for followers. They cannot rally around a leader if they do not know where he or she stands." [Garden (1990) 1. INTRODUCTION This article considers two key challenges facing leaders in large organizations: developing a successful mission for the organization and building high-performance teams. Developing a successful mission is a dynamic process that involves listening and incorporating new information about changes in the organization's environment. Team building involves aligning followers' incentives in order to facilitate coordination, information sharing, and the emergence of a productive corporate culture. Facilitating coordination is challenging because coordination is an activity that naturally has positive spill-overs. Thus, it typically benefits the organization more than it benefits the follower privately. Leadership can be a mechanism for resolving this incentive misalignment, if the leader can credibly commit to a course of action. The leader's dilemma isz REVIEW OF ECONOMIC STUDIES that he would like to base the organization's mission on all the relevant information about the environment available to him, But. since information about the environment trickles in over time. the leader may be led to revise the organization's direction as new information becomes available. His desire to modify the direction of the organization over time thus undermines his ability to coordinate actions and build highperformance teams. In this article. we consider how particular personal attributes such as steadiness or resoluteness help a leader to overcome this dilemom. The management literature on leadership has emphasized several key personality traits of good leaders. Among the most often mentioned are good communication skills. team spirit. integrity. and resoluteness. The lirst empirical study by economists that looks at the personal traits of leaders.-. considers which characteristics determine the professional success of CEO candidates involved in buyouts or venture capital transactions. Interestingly the study finds that. contrary to received wisdom that emphasizes the \"team player\" qualities ot' leaders. the traits that are the strongest predictors ot' success are execution skills and resoluteness. A general lesson from their study is that leaders should try to avoid changing direction over time and therefore should not seek too much feedback from others in the organization. Our model explains why resoluteness can he a desirable trait for a leader. and how it helps a leader in coordinating team actions. The model funher considers 'hottomup' information flows by exploring how followers may convey intormation by adapting their actions to the environment as they see it. If they expect the leader to pay attention to the information conveyed through their action choices then they will be induced to signal this information. while if they expect the leader to rely mostly on his own information then followers will give up on signalling through their actions and only worry about coordinating their actions with others. We suggest that the resulting multiple equilibria can be interpreted as different corporate cultures. Finally. the model explains why renegotiationproof incentive contracts that reward commitment to an initial plan of action cannot obviate the need for resolute leaders. More specilically. we capture the basic leadership problem in a simple setup involving four stages. In the first stage. the leader observes a signal of the environment the organization is likely to be in. Based on that signal. the leader can define a mission or overall strategy for the organization. Ina second stage. the other members ofthe organization. the followers. also observe signals about the state of nature and decide how closely they want to follow the leader's proposed strategy. They may not be inclined to blindly follow the leader's proposed strategy. because they know that in a third stage the leader receives a second signal and will only then commit to the organization's strategy based on all the information he has available. Thus. based on the signal they observe in the second stage. followers come up with different forecasts of what the leader's ultimate chosen direction for the organization will be. and coordinate their actions around their forecasts. Since by the third stage followers have already acted. the leader at this point is no longer concerned about coordinating their actions. The leader's only remaining goal is to adopt a strategy for the organization that is best given all the information he has. In the fourth and last stage. once the strategy has been implemented. the organization's payoff is realized. It will be higher the better adapted the strategy is to the enviromnent and the better coordinated all the members' actions are. The model considers a resolute leader who attaches an exaggerated information value to his initial information. or on the signals he processes himself. In other words. a resolute leader trusts his own initial judgement more than a rational leader would and discounts subsequently learned information. He therefore tends to define a strategy for the organization based disproportionately on his own best initial assessment of the environment the lirm linds itself in. The reason that such resoluteness is valuable is that the conflicting desires to coordinate followers and adapt the mission create a timeinconsistency problem. The leader would like {I o: '6 [pdv uo sopmqtj .{ltssastun etqurnmj 1e lo'S[t!lI.In0l_]:I.I0_]X0'P!'I1SJ.I :dltq Lung papemumoq BOLTON ETAL. LEADERSHIP, COORDINATION, AND CORPORATE CULTURE 3 followers to believe that his mission statement is what he will ultimately implement. But followers know that ex post the leader will want to revise the organization's strategy in response to new information after they have acted. This is what causes them to be insufficiently coordinated. as each attempts to guess how the leader will revise the organization's strategy in light of what they know about the environment. A resolute leader who puts too little weight on new information is more likely to follow through with the initial mission, which helps coordinate followers' actions around that mission. We show that this coordination benet outweighs the potential maladaptation cost as long as the leader's determination {or selfcondence) is not too extreme. Our model predicts that resoluteness is most valuable when the leader and followers are equally informed about the environment. When followers have little information, they have little reason to act differently from what the leader prescribes. Following the leader's direction. they coordinate closely. Likewise, when followers are very well informed, their assessments of the environment coincide and they also choose similar actions. It is inbetween. where coordination problems are most severe. that the value of a resolute leader is greatest. Thus one test of the theory could be to determine whether a leader's resoluteness {as measured by mm for example} has a humpshaped relationship with a measure of a leader's information advantage visavis followers. In the second part of the model, we combine both topdown and bottomup information flows by letting the leader's second signal take the form of an aggregate of followers' signals instead of an exogenous signal. In this variant of our model, the leader learns by observing followers' actions, which imperfectly convey their signals. In such a situation, letting followers base their actions on the signal they observe has more value for the organization, as this transmits more information about the state of the world to the leader. Since less coordination brings about better adaptation, observing actions moderates the benets of leader resoluteness. In this setting, resolute leaders make bad fisrcncrs and learn little, thereby destroying value. A leader's failure to listen to followers is especially costly when followers have very precise information. In this setting, our second main result is that observing followers' actions creates a feedback effect that can generate muftipfe eqm'fibrio: If followers expect the leader to ignore the information from their actions, then the leader will not learn anything new and his initial mission statement is the best estimate of the organization's nal action. Accordingly, when followers use the leader's announcement and not their private information to form actions, then the leader rightly ignores the aggregate action because it is uninformative. On the other hand, if followers expect the leader to listen carefully to the average action in revising the organization's policy, then they want to use their private signals to influence the organization's policy change through their actions. We suggest that an organization's corporate culture may determine which equilibrium and leadership styles prevail, so that our model can capture the hysteresis aspect of corporate culture emphasized in the management literature. Evidence on leadership styles in "collectivist" and \"individualist\" cultures supports our anal sis. In articular, the study of 20,000 managers and team members in 34 countries by Wenilt at if] ') shows that \"directive\" leadership styles, which leave less to the discretion of followers, are more prevalent and successful in collectivist than in individualistic societies. The remainder of our article is organized as follows. Section 2 reviews the literature. SectionEl presents our basic model ofcoordination and adaptation. Sectiong shows why leader resoluteness is valuable. Section 4.3 introduces a board of directors who can write an incentive contract with a rational leader. SectionE considers a more general model, where the leader can learn from the actions of others in the organization. Section concludes with a summary and directions for future research. Finally, an appendix contains the proofs and the foundations for our objective function. El 0;; '6 Ipdv no saumqtj alteration ctqummj 1c m'steumotho'tso'pnlsaJ,.-',.-':dllq we.\" popcotumqj 4 REVIEW OF ECONOMIC STUDIES 2. RELATED LITERATURE Leaders play many roles. While we focus on teambuilding and developing a mission, other studies examine the ability of a leader to motivate and communicate. Although some of these studies also feature a role for coordination of followers and leaders with distorted beliefs, three key ideas differentiate our study: {1} leadership is about overcoming a misalignment of followers1 coordination incentives; (2) the ability to renegotiate contracts creates commitment problems that personal characteristics of a leader can overcome; and (3) the personal characteristics of a leader affect the [low of information from followers to leaders and thus the culture of the organization. These ideas are motivated by a vast management literature, which often describes leaders' tasks as communicating with followers, coordinating their actions, and choosing a direction for the rml Furthermore, our exploration of personality traits as a key determinant of the leader's success is motivated by a vast body of evidencel In what follows, we contrast our framework with other models of leadershipl Leader as motivator: W m) model an empathic CEO, who gives weight to followers\" utilities. Leader em ath increases followers\" incentive to devise new ideas. When a lirm has two activities, m show that CEO bias in favour of the more romising activity provides incentives for agents to work harder on that activity. m considers a moral hazard in teams problem. He shows that a leader who exerts effort (leads by example) can signal that the return to effort is high, which motivates his team. Leader as communicator: In lDewan and Myatt -), followers would like to coordinate, but cannot because they do not know what others believe. Thus, the role of the leader is to communicate information that can facilitate coordinationE In contrast, our followers have an insufficient incentive to coordinate. Ourleader uses commitment to resolve this incentive problem, in a way that makes the outcome timeconsistent and does not inhibit the [low of information from followers to the leader. Leader overconlidenee: What we call resoluteness is similar to overcaaa'enre because both involve overestimating the precision of one's information. But overconfidence simply refers to people who overweight some information, without being specific about which information is overvalued. Resoluteness describes a manager who overweights his initial information. When the later information comes from others in our extended model, we can also interpret resoluteness as the tendency to discount information acquired from others. The overconfldence model most similar to ours is because managerial overconftdence also serves as a commitment device. In his model, that commitment helps attract and retain similarly minded employeesE _ The mana cmcnt literature on 'stratcgic leadership' discussed in m) is closest to our deElI | analysis. Barnard and lSelznichl d1952ll highlight the CEO's role in dening the rm's mission and fostering coordination. 2. Vision, judgement, charisma, resoluteness, as well as interpersonal, relational and communication skills are all commonly extolled. See ang. lPt'ct'fer and Salanikl ll'i'll) and lHambrick and Finkelsteinltl1939l]. Another branch of this literature. stresses behavioural aspects of strategic leadership, such as the leader's limited attention. as e. I. in m. The leader's ability to thrive on ambiguity and risk is emphasized in I _ I I I _ I ). The leader's ability to inspire followers. his charisma. and importantly. his selfcondence is widely emphasized as cg. in Basd l. 3. For a more extensive discussion of the economics literature on leadershi -, see 20] l . 4. Other studies in this s iirit are - 5. See also, m I r . where overconlidcnt managers win tournaments. or Blanes i Vidal and Moller Illllt} and Gervais and Goldstei I200 . where overcondent managers better motivate followers. 5mg '6 {pdv uo sopmqn Kitsaantun etqwmo 3 It: ,w'sicumopro}xo'pttlseJ,.-;.-':duq we.\" pap'eoprmoa BOLTON ETAL. LEADERSHIP, COORDINATION, AND CORPORATE CULTURE 5 Corporate culture: offers a relational-contract theory of corporate culture involving innitely lived rms and nitely lived workers, who must be given incentives to exert costly effort. We share Kreps' premise that corporate culture is related to endogenous equilibrium beliefs. But our idea is that corporate culture is about agents\" interactions with others, and the information that interaction generates. 3. MODEL SETUP The tension between coordination and exibility arises from the gradual arrival of infor mation about the environment. To illustrate this problem we consider a setting similar to Dessein and Santosl-), where the leader receives an exogenous signal in each of two periods. Based on his initial beliefs, the leader proposes a strategy for the organization around which other members can coordinate their actions. But the leader may change his mind and reorient the strategy following the arrival of the second signal. While the ex post reorientation helps bring about better adaptation, the anticipation of possible changes in strategy also make it harder to coordinate followers\" actions. The reason is that the followers also observe a private signal about the environment and use this signal to forecast possible reorientations of the organization's strategy] We show that leader resoluteness is a valuable attribute in such a situation (Section. The more resolute the leader the less likely he is to change his mind and therefore the less likely is a possible reorientation of the organization's strategy. We assume for now that signals are exogenous. We explore endogenous signals, derived from the aggregate choice of followers, in Section E Model setup: The organization we consider has one leader and a continuum of followers indexed by t'. The organization operates in an environment parameterized by 9, which affects payoffs. The better adapted the organization is to its environment the higher is its payoff. The difculty for the organization is that 6 is not known perfectly to any member. The leader of the organization and the other agents {the followers} start with different information or beliefs about the true value off). The leader differs from the followers in two ways: rst he can dene a rtrr'ssr'orr starertrerrt for the organization based on his initial beliefs 9L RAW-34.08). Indeed, leader's mission statement optimally reveals 6L. Followers make their own moves after seeing the mission statement; they have diffuse priors and also obtain their own private information about theenvironment, S,- =9 + 9,- (with i.i.d. error terms e;~.-V{0,a%}}l Second, after the followers have chosen their actions a,-, the leader receives further information about 9 in the form of a signal SL=9+L, where er; ~J'v"{0,of). This second signal (Si) can be either an exogenous signal or an endogenous signal, such as an average action, which reflects followers\" information. The fact that the leader gets more signals than followers is not important. Since followers\" actions have already been chosen at this stage, giving them a second signal would not change any results. What is important is that the leader makes the nal decision. We could even allow followers to adjust their action choices as new information arrives. But as long as there is some cost to taking actions before the decision date that are poorly aligned with the leader's decision, the followers will have an incentive to forecast the leader's choice, which is what spawns the coordination problem. 6. Th_ -) model does not consider an organization that is led by a leader. However. the basic tradeot't' our leader faces is similar to that in the organizational design problem they consider. The success of their organization like ours depends on coordinating many agents' tasks while adapting to a changing environment. "I. Note that we do not depart from the common prior assumption, which allows consistent welfare statements. One can think of the initial beliefs as resulting froth updating a flat (improper) prior based on an initial signal. gm: '6 Inch: uo saperqrj Mrsranrun erqurnro 3 re ,.-J0's[eurnopro'rxo'pmsaJ,-;.-':drrq wort papeorumoa 6 REVIEW OF ECONOMIC STUDIES Followers value three things: 1. belonging to a wellwcoordinated organization 2. taking an action that is aligned with the organization's strategy, and, 3. belonging to an organization that is well adapted to its environment 9. Formally, we represent these preferences with the following payoff function for each follower: r1,-= [{j 5)sz ta.- at? (a. 6)2forie[0.ll, {1) .t' ifw is the average followers' action. The payoff of the or anization, and its leader, is then the sum (in our case the integral) over all followers\" objectives That is, where n: jug 51>sz jun,- azdj (at. a)? {2) f f The leader's objective is to maximize ELfl'l], where EL denotes the expectation, given the leader's beliefs. The organization's objective is E [1'1], where E is the rational expectation under the ex ante probability measure. This objective function captures the essence of team-building: Since the social benet to coordination exceeds the private benet, there is a role for someone to enhance the team's welfare by encouraging coordination. When each follower is well coordinated (a; close to 5), the entire organization benets. But because each follower has zero mass, his personal benet from coordination is zero. Thus, there is a positive coordination externality. Appendix IE outlines some foundations for this objective function. The key elements of the foundational model are leamingby-doing and the requirement that wage contracts be renegotiation proof. Followers\" actions in this strategic interaction determine their value to the rm, and therefore their wage in future periods. This, in turn, prompts them to choose actions close to the ultimate direction of the rm aL, which have greater value. The requirement that wage contracts be renegotiation-proof precludes ex ante contracts that induce followers to coordinate efciently. This forms the basis for modelling the benets to coordination as a positive externality. Central to our analysis is the notion that the leader is resolute: he may overestimate the precision of his initial beliefs. More formally, although initial beliefs are truly drawn from a 8. Note that coordination does not have to mean taking identical actions. For ex ample, a leader might want to assign each follower to a different task b5. If I),- =f+a,-. then coordination {a;=a, Vt} would mean that actions are uniformly distributed over the unit interval. 9. A separate online technical appendix posted on the auLhors' websites explores alternative payoff formulations. Assuming the leader and the rm have different objectives so that the leader has no concern for misalignment leaves our qualitative conclusions unchanged. Similarly, if we weight the three terms of the payoff function unequally. it does not reverse our conclusions. A greater concern for alignment or coordination makes the optimal level of overconfidence higher. while a greater concern for adaptation makes it smaller, but still positive. Finally, the Appendix explores different forms of the coordination externality and commitment cost. 5] oz '6 {new uo separq Kttsrantun etqumjog 19 gJosIsumoprojch'pmsartgrthtq wort papeommog BOLTON ET AL. LEADERSHIP, COORDINATION, AND CORPORATE CULTURE 7 distribution OL ~N(0, 7) a resolute leader believes that they have a higher precision, 1/072 1/0 . In contrast, the second signal SL ~N(0, of) is assumed to be correctly assessed: we assume that the true and perceived precision of this signal are the same 19 The rationale for modelling resoluteness as a higher precision of the leader's initial belief, is most clear in Section 5) when the signal SL is generated by other agents' actions. In essence, resoluteness in our model means that a leader trusts his own judgement more than the information acquired from others. But for now, the leader cannot observe followers' actions or signals. 4. MERITS OF RESOLUTENESS We begin by analysing the case where the leader's second signal, SL, is exogenous, and solve for a unique linear Perfect Bayesian Nash Equilibrium of the game described above ! ! Definition 1. A Perfect Bayesian Nash Equilibrium is given by Downloaded from http://re (i) a strategy, or direction, for the organization al that maximizes E[ILIOL, SL], given followers' actions tailieto, 1]; (ii) followers' actions a; that maximize E[II;IOL, Si] given at and (a; ]jero, 11' (iii) Bayesian updating: ELOIOL, Sil=$FOL +(1-OF)S; where OF 210 and EL[OIL, SL]=$LOL + (1 -QL)SL where OL = 2 Optimal actions: We solve the model by backwards induction. When the leader chooses the organization's strategy aL, the actions of the followers failieto, 1] are already determined. We will guess and verify that the leader chooses an action that is a linear combination of his two signals: aL =aLOL+(1-aL)SL. (3) Knowing this action rule, each follower i chooses an action a;. For simplicity we assume that followers start with a diffuse prior, which they update using the leader's mission statement Oz and the signal Si ~N(0, of) they each privately and independently receive. Any follower takes the actions of the others as given and cannot influence the average action because he is of measure zero. Therefore, his objective function (1) reduces to E [-(a; -aL)210L, S;] and his optimal action a; is equal to his expectation of the leader's action, given his own private signal S;: Libraries on April 9, 2013 a; = E[aLleL, Si]=aLOL + (1-aL)E[SLIOL, Si]. (4) Since Sy is an independent, unbiased signal about 0, E[SLIOL, Si]= ELOIOL, Si]. By Bayes' law, the followers' expectation of 0 is E[OIL, Si]=OFOL + (1-OF)S; where 0- 2 - 2 +OF 10. Note that if we allow for costly information acquisition by the leader at date =2 then our model allows for an alternative interpretation than leader overconfidence. If the leader under-invests in information acquisition-as he would if he privately bears all the costs-and if this is observable (or anticipated) by followers when they act, then under-investment in second period information will have the same effect as resoluteness in our model: the leader will put more weight on the first signal. 11. Our definition implicitly assumes the trivial result that the leader's initial mission statement coincides with er.8 REVIEW OF ECONOMIC STUDIES Let aF denote the weight that the follower puts on the leader's announcement when forming his action. Then, ai = aFOL + (1-QF)Si. (5) where 1-aF= (1-aL)(1 -4F). (6) Now, we use the knowledge of followers' strategies to determine the leader's optimal action. The first-order condition of the leader's utility function with respect to al is Rearranging, aL = (Ella]+ EL), (7) where a =fajdj is the average followers' action and Ez denotes the expectation, conditional on the leader's information set at the time when he chooses his action. That information set includes his initial belief Oz and his signal SL. The leader's expectation of the state 0 is given by Bayes' law: EL[0]=$LOL + (1-QL)SL (8) where QL:= or or - 2 +0 1 To determine the average follower's action, integrate over (5), noting that the mean of the follower's signals is the true state 0. Thus, a=aFOL + (1-QF)0. The leader's expectation of this average action a is Ella] =QFOL +(1-OF)($LOL +(1-QL)SL). Substituting the leader's expectations into his optimal action rule (7) yields aL = ( OFOL + (2 -OF) ($LOL+ (1-QL)SL)). (9) Finally, collecting coefficients on OL then implies that aL = (OF + DL(2-QF))/2. We now know how the leader will act in equilibrium, given how the followers act. But this expression still has an unknown coefficient: aF. To solve for this coefficient, we substitute out OF from (6) and then solve for aL. This reveals that the leader's optimal action is in fact linear, of the form in (3), where 1 - L OL = 1 - 1 +QL+$F(1 -QL) (10) Using the relationship between or and of in (6) and rearranging terms reveals that followers' actions are also linear in OL and Si, with OF =2 1 - 1 1 1 + dL + OF(1-PL)) (11) Optimal resoluteness: Just like the leader's payoff, the organization's payoff II has three components. By substituting in the optimal actions of the leader and the followers, we can evaluateBOLTON ET AL. LEADERSHIP, COORDINATION, AND CORPORATE CULTURE 9 the effects of leader resoluteness and determine the optimal level of resoluteness. Following substitution of the equilibrium actions ar and (a;lieto, 1], the three components of II are as given below: 1. the variance of each follower's action around the leader's, [J,-(ar-a2)-di =-(1-a1)2 [progto;]. (12) 2. the dispersion of followers' actions around the mean, 1- (a;-a)? dj=-(1-a1)2(1-OF)$Fog, (13) 3. the distance of the leader's action from the true state, E[-(aL -0)2] =-(1-@1)202-0703. (14) Summing the three terms, substituting in (6) and using the definition of of then yields, En=-(1-a1)2 034F(2-DF)+202] -azo3. (15) Note that the effect of resoluteness appears in (15) only through the weight or that the leader puts on his first signal (OF, 60 and or are exogenous). Differentiating (10) with respect to L, and, in turn, differentiating or with respect to of, then reveals that day/do, 2>0. In other words, Downloaded from http://restud.oxfordjournals.org/ at Columbia University Libraries on April 9, 2013 aL is monotonically increasing in resoluteness. Therefore a simple way of determining the effect of leader resoluteness on the organization's welfare is to differentiate the ex ante objective with respect to oL. The chain rule then tells us that ElI/do, has the same sign. The partial derivative of the organization's ex ante expected payoff with respect to aL is: aEn aaL =2(1 -aL) OF(2-OF)03+ 207] -20108 This is positive if $F (2-OF) +2- 1 -aL Since aL/(1-aL)=OF+204, the above inequality becomes OF(1-OF) 2 With a rational leader we have of"=0 , and therefore the above inequality then reduces to OF(1 -OF) > 0 which always holds for of 0, which implies that delt > 0 at 0,2=00 , so that some degree of resoluteness is always optimal. On the other hand, for an extremely resolute leader who fails to update at all, o, - 0, d1 - 1, OL - 1, and the right side of the inequality approaches infinity, so that 90 0. The study by Wendt et all (2009) supports this prediction. They find that, in collectivist societies, "directive" leaders are more prevalent and are reported to be more successful than in individualist societies. The directive leaders described in the Wendt et all (2009) study are similar to our resolute leaders in that they exert more influence over their followers' actionsBOLTON ETAL. LEADERSHIP, COORDINATION. AND CORPORATE CULTURE I] and they tend to suppress discussion and dissent, a point that will be even more obvious after introducing a bottom-up information ow in Section- 4.2. Other ieaciership traits Are there other behavioural traits of leaders besides resoluteness that might be of value? Some of the most frequently studied behavioural biases in economics are eonrmcirow bias anchoring, m i o rimism lack ofsel leonrml or resent bias. and rational inattention (see e.g. nml m.m m. It is possible to explore the consequences of conrmatory or present biases, anchoring. or rational inattention by the leader in our model. Conrmatory bias can be viewed as a psychological underpinning of resoluteness: a leader who tends to put more weight on signals that conrm. and less weight on signals that contradict his prior, essentially acts like a resolute leader in our model. Similarly. a leader who anchors his beliefs around his initial signal helps improve coordination among followers. Optimism. on the other hand, biases both the leader's mission statement and nal choice in a particular direction. Followers. knowing that their leader has biased beliefs, will adjust their actions accordingly thus leaving coordination and alignment almost unchanged. The overall effect of the leader's optimism. however, is that the organization is on average less well adapted to its environment (and the loss term in the objective functionEI(aL6)3]is higher). A leader's present bias also generates inefciencies. Present bias arises when there is a misalignment between the rate of time preference today and the rate of time preference in the future. This misalignment can drive the leader to make a mission statement that is optimal given current objectives. but that is not always utility-maximizing for the leader in the future. This time- inconsistency problem for the leader exacerbates the followers' challenge of second-guessing the leader's ultimate choice of action and therefore undermines the leader's ability to coordinate followers. Finally, a leader's rational inattention can also provide an underpinning to strengthen the leader's resoluteness. Rational inattention is a form of bounded rationality initially proposed in@ m). Rationally inattentive agents observe information through limited bandwidth channels that add idiosyncratic noise to every signal they observe. They choose how to allocate their attention. which amounts to choosing which signals they want to observe more precisely. . . . . . . _ _') There is a cost to paying more attention, which in our model can take the form do\" 2'\"L \"l. Suppose further that e(.) is symmetric (clan2.014)=e{rr,:\".rru_2)) and increasing and convex in both arguments. Under these conditions. it is optimal fora leader to pay more attention in the . . . . . . . _ _"} initial stage and less attention to subsequent inlormation {on 2 2:- or'L \"). . . . _"} _'} . _ Praposrllon 3. if a leader can choose orU " and \"i. \". subject to a cost C(au 2 . . . . . . _') _') symmetric. increasing and convex. then a is optimal to choose an " :- \"L ". .aLZ). rhar is Paying more attention to the initial information helps remedy the organizations' coordination problem because it provides clearer information to followers before they need to choose their actions. Furthermore. getting more precise information initially reduces the marginal value of acquiring more information later. Thus, it also acts like a commitment device to deter future information acquisition and future changes in actions. When the inattentive leader updates his beliefs. he will place very little weight on the second, noisier. signal and therefore more weight on the rst signal. By weighting the first signal more. this leader will end up choosing an action close to the initial announcement. just like the resolute leader. oz '5 [pdv an $19qu igsnsrun etqulmejie .m'qeumofpmjxo'pnisar :diiq umJj papeonimoq [2 REVIEW OF ECONOMIC STUDIES 4.3. Renegminrmproof contracts cannot subslirmclfar resoluteness As the preceding analysis highlights, resoluteness of a leader provides a form of commitment to staying the course. It ensures that the leader's strategy choice after learning new information does not deviate too much from the mission he set for the organization, which is centered on his initial belief. If the leader's beliefs do not change much. his strategy choice will be similar to his mission statement. This commitment in turn facilitates coordination. However. to the extent that leader resoluteness also introduces a bias in the organization's adaptation to the environment. it would seem that a more direct solution to the leader's time~ineonsistency problemallowing a rational leader to commit to staying the course, or writing a contract that penalized the leader for a lack of commitmentwould be preferable. The question is: Why can not the organization simply write an optimal contract with a rational leader. instead of choosing an irrational one? The general answer is that it is not possible to achieve optimal commitment through a renegotiationproofcontract. In fact. it can be shown that the organization has a timeinconsistency problem for the same reason the leader does. It wants to commit to a penalty for action choices a; that deviate from 9;, but after followers have chosen their actions and coordination is achieved. the organization would like to undo the penalty to allow the leader to choose an action closer to the true state so that the cost of maladaptation (\"L 9}2 is also reduced. Moreover. the reduction in the penalty at that point also reduces the size of the deadweight loss from the transfer. Thus. by renegotiating the contract. the leader and the organization can reduce the maladaptation cost and the deadweight loss and make themselves both better off. If such renegotiation is anticipated. of course. followers will not pay attention to the leader's compensation contract. with the consequence that the organization will suffer from too much miscoordination under a rational leader. Note also that any contractual commitment over and above the commitment through the leader's resoluteness would be ineffective for the same reason. Followers would anticipate renegotiation. and would therefore simply ignore the leader's incentive compensation. 5. LEARNING FROM FOLLOWERS AND CORPORATE CULTURE In this section. not only do followers learn from their leader (topwdown information flow). but leaders may also learn from followers (bottomup information ow}. Accordingly. we now replace the exogenous signal 3L with an endogenous signal. which is the average action of the followers. plus some noise. A rst implication of the introduction of a bottom~up information channel is that this could moderate the benet of resoluteness: a leader who is very stubborn dissuades his followers from acting based on their private information and suppresses information revelation. More interestingly. because the leader's action depends on what he learns from agents' actions. which in turn depend on what agents expect the leader to do. multiple cquilibria arise. which can be interpreted as different outcomes arising from different corporate cultures. 5.1. Merits and drawbacks of resoluteness Suppose that followers' actions aggregate into the second signal for the leader. which now is the publicly observable organization output A: A =fajdj+em .5 where m is the independent noise term 8,1 ~JV(0.U). but that otherwise the model is the same as before. The leader then uses the signal A to update his initial belief 9!. and make a nal El 03 '5 [tJdv no saumqtj misusing etqumwj 1e 5.10'S[cumoprOjXO'pl'lEalf dnq Loon papeommoa BOLTON ETAL. LEADERSHIP, COORDINATION, AND CORPORATE CULTURE l3 inference about 9 . As we shall verify, followers\" equilibrium strategies again take the form a;(S,-} = ap9L+(l or_r}.S',-, so that we can rewrite the aggregate output signal as Str- ' [AO'FSL]=9+ ' 9,1. ltIF l[X'F Note that this signal's precision is given by {1 rxp)3crA_2, so that the more followers rely on their private information (the lower is our}, the more accurate this signal becomes. Of course, if followers rely more on their private signals S; there is also less coordination among them. Thus, in this setting coordinated actions have both a positive payoffextemality and a negative information externality because they suppress information revelation to the leader. Optimal actions: The nature of the leader's problem has not changed. As in Section ii the leader's optimal action is aL=E[9|9L.L]=OIL9L+(1-&'L}t. where or; is given by Equation .. That solution for on, is expressed as a function of m, the weight the leader puts on 9,: when updating her belief about 9 using Bayes' law. The change to an endogenous signal, A, shows up as a different 4'); from before: 072 >L='. {18) Ur_1+(1-F}20A_2 The difference in this case is that 425; now depends on up, which is chosen by the followers and will, in turn, depend on the leader's resoluteness Cir2. Similarly, each follower's optimal action is their forecast of the leader's action, which can still be expressed as f=QF9L+(l th}S,-, where (2;: is given by Equation dB. However, the difference is again that now or; depends on (35L. Thus, in this setting with bottomup information flows, all depends on or; and conversely or; depends on tag, so that we now need to solve the fixed point problem given by Equations and (1m to determine the equilibrium actions of leader and followers. Substituting for (pg in Equation 11m delivers a thirdorder polynomial in (1 uIF}: (1ar}[t1a.:}2o,;3t1+r}t1ar}o;2(1F}+2a:2]=0. This equation potentially has three solutions. or}: = 1 is always a solution, for any set of parameter values. The quadratic term in brackets also has two zeros if Hz that? r _ . {19) so +moj Since we focus on stable equilibria we neglect the unstable equilibrium with the larger quadratic root for Up. The following proposition characterizes the two stable solutions. Proposition 4. When leaders learn fmm followers' rtetirms wholdsj, there are two stahle (linear) equilibria: ( i J A dictatorial equilibrium where there is perfect coordination a]; = a; = u but intrmation owfmm followers to leaders is totally suppressed. El 0: '6 {pdv uo ssumq misusing eiqtunlo 3 It: sJO'SIBU'mOprQO'plSStffidllq LL10.\" papeommoq 14 REVIEW OF ECONOMIC STUDIES (ii) A \"lead-by-being-ied equilibrium\" where coordination is reduced, but the organization is better adapted to the environment. as it relies on more infomation to determine its strategy: 2 2 2 1-F+'l(1-F} -3(1+F}0r 0;, a5=th9L+(l o:,r)Sg where crp=l 2(1 +Fl (30) and . 1F+,/(1F}28t1+na:2aj aL=QL9L+(l&L)SL where 0:1,: 1 2(1%) (21) The economic logic of the multiple equilibria is the following: If followers expect the leader to ignore any new information from their actions, then they must also expect the leader's action to be the same as the organization's mission statement (aL=9L}. Since followers want to take actions close to the leader's action, they then choose the same action a; = 9:4. But when followers all take the same actions, they reveal no new information. So, their expectation is selfconrming. In contrast, when followers expect the leader to learn new information from the observed output A, they try to forecast what he will learn, using their private signals. Because their actions are based on this forecast and on their private signals, aggreate output reveals information. So, the expectation that the leader will learn is also conrmedi 5.2. Matching leaders and corporate enitnres One way of interpreting the multiplicity of equilibria in this setting is that the role of leadership in an organization must be adapted to the organization's culture. There is no point in assigning a leader that is a good listener in an organization that has a hierarchical and dictatorial culture. Viceversa, appointing a very resolute leader in a democratic organization in an attempt to bring about greater coordination could be costly, as this may clash with followers' incentives to take initiatives. These observations have often been made and are well understood in the management literature. Resoluteness and organizational aligmnent: A somewhat unexpected prediction of our analysis is that the assignment of leaders with different degrees of resoluteness to organizations with 12. This multiplicity ofequilibria is generally robust tothe introduction ofa second exogenous signal. To see this, let the precision ofthis exogenous signal be denoted 5. Then, the total signal precision ofthe leader's second signal is \"it2\" err)2 +5. Equations {5} and (II) still characterize followers' optimal actions, while (3] and {ID} characterize leaders' optima] actions. The only difference is in the denition of the Bayesian updating weight qu. Now, L=rrf2ftafl+ of\" ar_;.-)2 +6. Substituting the new denition of 135;, in (Ill) yields a new expression for {l 0-'L}- Substituting this new expression for {l QL) into (I a;) =(l a;_)tl J,-] and rearranging yields, u w->[u 'r'}10,.1_2\"+rlU whirl rl+2cr.._2+(l was] (Ie-)s=u. Even in the presence of an exogenous signal {5 ; ll), this is still a cubic equation in (l a,-) and it remains cubic as long as the leader receives some information from an endogenous signal. Therefore, multiple equilibria do not disappear with the introduction of an exogenous signal. Instead, it is the unique, linear, equilibrium {obtained when the leader only receives an exogenous signal) that is fragile. As 5 approaches zero. there is a solution on: that is close to I. So. with a small amount of exogenous information, there is still a corporate culture that is very dictatorial, with very little listening. gm: '6 {trdv uo sapmqt'] Romanian etqwnIOJ It! ,Josmumo[pJo'tch'pmsai-'i':duq wort papeotumoa BOLTON ETAL. LEADERSHIP, COORDINATION, AND CORPORATE CULTURE 15 different corporate cultures is not a simple matter of matching more resolute leaders with poorly aligned organizations. Instead, it is the organizations where followers' actions are already well aligned with the leader's action where resoluteness is most valuable. In organizations where the noise in the leader's signal 0,1 is low, the leader will respond strongly to the signal, choosing an action that is close to the average agents' actions. As a result, the leader's actions and followers actions will be closely aligned. Thus, we vary 0,; to better understand the culture of alignment in the organization and how it corresponds to the optimal characteristics of a leader. In the dictatorial equilibrium (03;: = l, or; = 1}, leader resoluteness has no effect on the organization's ex ante expected payoff because, in this case, the coefcients aL, asp and $1, do not depend on the leader's resoluteness. In the stable leadbybeingled equilibrium, on the other hand, the organization's expected payoff is 2'51? 2 7 m \"LUG- Leader resoluteness affects this payoff through 0:1,, the weight the leader puts on her initial signal 61, when choosing her action. The relationship between the weight 0:1, and resoluteness 0,72 is described in the second equation of Proposition. Taking the derivative of the payoff with respect to org, and setting it equal to zero {BEN/8a,; = 0), yields the optimal weight that the rm would want a leader to put on his initial signal: en: (1 az)'F{2F)oi F(2'F} . {22) 1+F(2F} * CEL This is the optimal weight as long as the leadbybeingled equilibrium exists and the secondorder condition holds. The existence condition can be satised if the noise in output, the degree of leader resoluteness, and the true precision of the leader's initial belief are low, and the precision of agents' private information is high. More precisely, we are able to show that: Proposition 5. In the Iead-by-being-iea' equilibrium, ieaa'er resoluteness increases the organi- zation '1: expected payoff if and (mi); if 02 brocar 2 A C 2t1+p(2F2' {23) Otherwise, the opposite of resoiiiteness. "exibility", increases the expected payo When is the leader's resoluteness likely to be benecial? It is, for one, in situations where the leader is already extracting most of the relevant information about the environment 9. If the signal the leader sees from the followers' output is already very precise {low of ), then the benet of better coordination (F(2 pp matters more than the marginal loss of signal quality. When the leader learns little from followers' actions (of is large), then somewhat surprisingly, leader resoluteness worsens coordination problems in the lead-bybeing-led equilibrium. The leader's action is then not very responsive to the signal A, thus resulting in more miscoordination. In such a situation it may actually be preferable to have a exible or acquiescent leader. Leader acquiescence then induces more initiatives from followers, which in turn allows the leader to observe more precise information about 9 and take a betterdirected nal action. In sum, resoluteness is most valuable when coordination is highly valued, there is little noise in output and the true variance of the leader's initial belief is high. 5] oz '6 {radio uo SDIJEJQI'I ttsrantun etqumto 3 re I-JO'SIEUJHOFPJO'IXO'pnls-DJI--'idllq wort papeotumog 16 REVIEW OF ECONOMIC STUDIES 5.3. Ranking equilibria.' can resoluteness be preferable to competence." Given that there may be multiple equilibria and that corporate culture may therefore matter, is it possible to say which corporate cultures are better? Is a dictatorial culture always better given that it leads to better coordination? Or can a democratic culture (under the lead-by-being- led equilibrium} bring about better performance due to the organization's greater adaptability to changed circumstances? The next proposition provides a clear ranking of the dictatorial and optimal lead-by-being-led equilibrium {that is, the lead-by-being-led equilibrium with an optimally resolute leader). It establishes that if the endogenous second signal SL is sufciently informative (org? is low enough) then an organization with a democratic culture dominates a dictatorial organization. Proposition 6. The optimal leadbybeingled equilibrium exists if condition holds and it dominates the dictatorial equilibrium if and only if: 2 (1 ttvr)2 2 . 24 0,1: 2t1+pt2rn"\" ( ) Allowing for different values of 06' is a simple way of introducing differences in a leader's competence into our model. A highly competent leader then would be one who has a highly accurate initial belief 91;, that is someone with a low value of 002. Intuitively, one expects greater competence of a leader to be an unreserved benet for an organization. A leader with more accurate initial information, would make better decisions other things equal, and this can only benet the organization. However, greater competence of a leader in our model also has a side effect: it may crowd out learning from the actions of followers. If the leader's initial information is too precise he may no longer be able to learn anything from the actions of the followers, as the latter decide to ignore their own information when choosing their actions. The question then arises whether it may be preferable for the organization to have a resolute leader who knows less, but who is also able to learn from followers. Our analysis has another surprising implication. We provide a set of conditions below on the parameters of the model such that the organization is better off with a resolute leader rather than a more competent, rational leader. Such a situation may arise when it is better for the organization if the leader learns from the actions of followers, and when only the resolute leader is able to do so in equilibrium. Observe rst that when 05 varies, the resolute leader's Bayesian updating weight m is unaffected, as the leader believes the variance to be 0,93. The followers' beliefs are affected because when the true precision of the leader's announcement 002 changes. the weight followers put on that announcement when forming expectations of the state becomes (,'lJF =ao_2/{0F +00 Given this new expression of dip, the leaders\" and followers\" actions take the same form as before. Thus, the solution is again that given by Proposition 2, and the lead-by-being-lead equilibrium exists whenever {E25 holds. In sum, changing competence only affects the solution through its effect on the value of (by. We are then able to establish the following proposition. Proposition 7. Suppose there are two leaders. one resolute and one rational. Both have initial beliefs with the same perceived precision of 2 = 002 , but the resolute leader's initial information . . _9 _ . has lower true pre cision on \"