JohnAhlquist-Leadership What It Means

PL14CH01-Levi ARI 4 November 2010 11:1

RE V I E W

S

IN

AD V A

NC

E

Leadership: What It Means,What It Does, and What WeWant to Know About ItJohn S. Ahlquist1,3 and Margaret Levi2,3

1Department of Political Science, Florida State University, Tallahassee,Florida 32306-2230; email: [email protected] of Political Science, University of Washington, Seattle, Washington 98195;3United States Studies Center at the University of Sydney, New South Wales 2006,Australia; email: [email protected]

Annu. Rev. Polit. Sci. 2011. 14:1–24

The Annual Review of Political Science is online atpolisci.annualreviews.org

This article’s doi:10.1146/annurev-polisci-042409-152654

Copyright c© 2011 by Annual Reviews.All rights reserved

1094-2939/11/0615-0001$20.00

Keywords

collective action, coordination, communication, organization,hierarchy, political economy, experimental methods

Abstract

Leaders are part of virtually all organized political life. There have beenimportant recent advances in modeling “leaders” as well as clever andinnovative empirical studies. We review recent contributions from thepolitical science, economics, and management literatures. We discussthe extent to which these new works represent advances over the majorclassic works on leadership and organization from the twentieth century.We identify important gaps, chief among them (a) theorizing a rolefor coercion, (b) modeling the endogenous emergence of leaders, and(c) empirically disentangling the effect of an individual leader from heroffice, especially when leaders emerge endogenously.

1


INTRODUCTION

Heads of organizations—be they popes, pres-idents, generals, CEOs, or general secretaries—must coordinate their followers to producedesired actions and outcomes. Their tools in-clude coercion, incentives, and persuasion, butin what combination and to what effect leadersrely on these tools are variable.

We also observe substantial variation inthe behavior of members and the outputs oforganizations. Some leaders appear able toelicit greater effort or sacrifice from their fol-lowers than others. Within the same industry,workers in some firms are demonstrably moreproductive and committed than in others.Some governments can raise armies and taxeswith a minimum of coercion while othersrequire considerable force and policing to gaincompliance. Political parties and voluntaryorganizations vary greatly in the level ofactivism and membership loyalty. Some socialmovements succeed while others fail miserably.

Observers often lay credit or blame at thefeet of organizational leaders for marked shiftsin aggregate behavior within polities and or-ganizations. The role of leadership in effectingor blocking change is a major emphasis in cur-rent research on development, reform, and thesolution of common pool resource problems.Certainly history is often presented as a seriesof stories about certain individuals cast as “ar-chitects of political change” (Schofield 2006)able to transform beliefs about what is possi-ble and consequently alter behavior of largeproportions of a society. Illustrations includeblack leadership in the United States (Marable1998), social movement leadership worldwide(McAdam et al. 2001, Tarrow 1994), and busi-ness management (Heifetz et al. 2009).

But how much of the variation in groupbehavior is attributable to specific leaders? Isthere an analytical distinction to be made be-tween a leader and someone who occupies anofficial position? Should we be focusing our at-tention on the office-holder or the office? Howdo we even determine who possesses leader-ship power? Is it possible to make systematic

statements about the factors that make lead-ers more or less effective at achieving organi-zational goals? Can there be authoritative ac-counts of when the role of leaders is minimalrelative to structural factors or political op-portunities? If indeed the nature of leadershiphelps account for variation in aggregate behav-ior, what then are the determinative qualitiesof leaders in those instances where leaders per-suade followers to change their course of actionor their positions on political issues, or to makecontributions beyond what other leaders seemable to evoke? Can we formalize any of this? Isleadership success a consequence of how leaderstreat followers, the information and coordina-tion they provide, their capacity to inspire, orsomething else altogether?

Leadership as an object of study in politicalscience and political economy tends to be cycli-cal ( Jones 1989). It has again come into fash-ion. Building on a long and multidisciplinaryresearch program on leadership, a substantialbody of work worthy of review has emergedover the past two decades. Some of the mostsignificant recent advances derive from politicaleconomic discussions of agenda control, coali-tions, the formation of beliefs and preferences,and information flows. Whether the focus ison the qualities of judgment and communica-tion essential for effective leadership (Dewan& Myatt 2008, Humphreys et al. 2006) or onthe role of leadership in building governmentsand effecting institutional change (Levi 2006,Schofield 2006), the recent literature draws onold and new models and methods to addresscritical questions.

At one level, much of political science isabout how people select, remove, and holdaccountable those empowered to make impor-tant decisions (e.g., Bueno de Mesquita et al.2003, Przeworski et al. 1999). Here we are lessconcerned with the rules affecting selection ofand constraints on elected leaders and bureau-crats than with the role of specific individualsin directing and coordinating followers in awide array of formal and typically hierarchicalorganizations. We explore how individualscome to be recognized as leaders (whether

2 Ahlquist · Levi


in official positions or not), when they arelikely to use coercion, and what they can getfollowers to do. Our focus is more general thanpolitical officialdom but not as broad as theforms of leadership addressed in the large andprimarily ethnographic literature originatingin anthropology, history, and sociology.

Innovative empirical strategies increasinglyconfirm the causal relevance of a leader, but thetheoretical understanding of the causal impor-tance of leadership is still fairly impoverished.The concept remains vague or contested, whichmeans the literature is incoherent; scholars of-ten use the same term to mean distinct actorsor processes. A divide continues between re-searchers who take the leaders as given and areinterested in the differences across them andthose more interested in how leaders emerge.We discuss these issues, identify promising newtheoretical work and clever empirical strategies,and outline gaps in the existing literature andopportunities for future research.

TAKE ME TO YOUR LEADER?

Classical research on leadership emphasizedpower and legitimacy (Burns 1978, Machiavelli1950 [1532], Weber 1968); the “functions of theexecutive” (Barnard 1938), that is, appropriategoals and how best to achieve them; the inter-actions of leaders with staff and followers (Blau1955, Burns 1978, Selznick 1957, Weber 1968);or the pathologies associated with leaders’ ex-ercise of power (Michels 1962 [1919], Merton1968 [1957]). Much of this work is on admin-istrative leadership, and the best of it detailsmore than staffing hierarchy or the formal re-sponsibilities of various members of the firm oragency. The authors also describe the political,economic, and regulatory context (Blau 1955,Simon 1947, Weber 1968). That context mat-ters is a truism; leaders are certainly constrainedby their position in history, as so many politicaland corporate memoirs emphasize. The trick isto clarify the relationship between context andthe particular attributes and tasks of leaders.

This vast literature offers some definitionaland analytic guidelines and even some findings.

The capacity to direct the actions of othersdefines and is a necessary condition of leader-ship, but it is only one of several characteristicsleaders may possess and even require. Leadershave power over others as a consequence ofoffice, personal influence, persuasive capacity,charisma, or coercion, but all exercise power ininteraction with followers. Without followerswho act on leadership directives, the title ofleader is hollow. It is followers who grantauthority and legitimacy, the sine qua nonof those who can evoke, not simply coerce,desired behaviors (Coleman 1980, Levi 1997).Some leaders tightly delimit the behavior offollowers; others expect compliance with a verylimited set of dictates. But in all instances, lead-ership implies the existence of an organizationin which a population responds to the leader’sguidance.

The emphasis on followers as key to a modelof leadership dates back at least to Machiavelli.Whether it is better for a leader to be loved orfeared (but definitely not hated) (Machiavelli1950 [1532], Ch. 7) continues to punctuate thedebate (Majumdar & Mukand 2010). Weber’stypology of legitimacy is simultaneously atypology of the grounds on which leaders haveauthority, exercise influence, and can demanddeference (Coleman 1963, 1980; Dahl 1957;Parsons 1963a,b; Tilly 2005; Weber 1968).Fear and love may be a consequence of dom-inance, but they subside in importance relativeto the bases for dominance provided by thesocial structure and political processes. Amongthe tools of dominance, as both Machiavelliand Weber insist, is coercion; it is not alwaysused but is nearly always available to those withleadership power.

Yet another perspective comes from Michels(1962 [1919]), who offers a psychology of fol-lowers; they are easily led and manipulated,which gives the leaders power to transform thevery goals and ends of the organization itself.Michels’ view certainly resonates in some of theinfluential post–World War II work on lead-ership, in which researchers sought explana-tions for why so many collaborated with theNazis and other brutal regimes. The important

www.annualreviews.org • Leadership 3


Milgram experiments on obedience (Milgram1974) and the research program on the authori-tarian personality (Adorno 1950) exemplify thispreoccupation.

A third perspective emphasizes the ex-changes that take place between hierarchi-cal superiors and subordinates and, indeed,among virtually all members of the relevantgroup (Blau 1964, Emerson 1976, Homans1958, Thibaut & Kelley 1959), a view that res-onates more with Machiavelli and Weber thanwith Michels. The emphasis on social networksamong members of an organization and the re-quirement that the leaders instill norms andcommitments within the group (Barnard 1938,Selznick 1957) continue to inform contempo-rary research on strategic interactions betweenleaders and followers. A long tradition of stud-ies of industrial organization (see summariesin Miller 1992; Cook et al. 2005, Ch. 7) anda more recent experimental literature (Fehr &Falk 1999, Fehr et al. 1998) suggest that whensubordinates are treated with respect, they re-spond with loyalty and greater productivity.Another outgrowth of this tradition is the workon brokerage, in which key actors facilitate rela-tionships among the participants in a work site(Brehm & Gates 2008) or in a collective action(McAdam et al. 2001).

Recent research in political science andeconomics builds on this earlier work, of-ten by formalizing ideas found in Machiavelli,Michels, and Barnard. However, the emphasison principal–agent and social-exchange mod-els has contributed to a neglect of the role ofcoercion, an issue to which we return below.

Scholars in other traditions also continue towrestle with the subject of leadership. Amongthem are political theorists (Keohane 2010,Walzer 1973, Williams 1981), scholars of theexecutive head of state (for a review, seeGoethals 2005), and management theorists (fora review, see Dirks & Skarlicki 2004). Psychol-ogists have been particularly active in the fieldof leadership studies (Messick & Kramer 2004,Avolio et al. 2009), illuminating cognitive lim-its and the nature of the interactions betweenfollowership and leadership.

In addition, the popular business press gen-erates thousands of titles on the topic of leader-ship. Many have the form of self-improvementmanuals, offering advice on how to motivateothers. A more likely source of insights on lead-ership is memoirs, biographies and, sometimes,hagiographies. The best of this work distilsfrom these stories the qualities and attributesthat make leaders effective in coordinating, de-limiting, and inspiring action.

More closely related to the social scienceenterprise are the papers and cases producedby business and public policy school facultyin their role as management consultants.Some focus on business enterprises and otherson political and economic development andreform efforts throughout the world. Thegoal of the World Bank’s Global LeadershipInitiative is “publication of a set of case studieson past and present capacity developmentinterventions targeting high-level strate-gic leadership” (http://www.capacity.org/en/resource_corners/leadership/research/global_leadership_initiative_of_the_world_bank_institute). Cases can prove very usefulas grist for developing or testing theory.Unfortunately, most of this work tends to bedescriptive and taxonomic. It offers “just so”stories that are at best no more than a statementof intelligent or common-sense intuitions. Theworst of the work in these traditions is simplyad hoc and glib.

A review of the social science literaturereveals an equal dissensus on the character-ization of leaders or leadership. For certainauthors, a leader is a pivotal individual whoembodies and articulates a model of the world(Henrich & Gil-White 2001). Some emphasizethe charisma of leaders (Brown & Trevino2009, Rowold & Laukamp 2009, Tucker 1968)or their transformational capacities (Avolioet al. 2009). Others describe leaders as thosewho communicate clearly (Dewan & Myatt2008), manage interpersonal relationships, andset up incentive systems to solve coordinationproblems of a group. Leaders may be thosewho have secured a position of power andinfluence but who make difficult or unpopular

4 Ahlquist · Levi


decisions (Canes-Wrone et al. 2001). Finally,“leader” might simply be a synonym for thetitular head of some hierarchical structure.Suffice to say that if the little green men usedthis literature to prepare themselves for Earthlanding, they would not have a clue aboutwhom they would meet upon demanding,“Take me to your leader.”

TOWARD A MOREINTEGRATIVE APPROACH

The seeming discordance in the literature maybe misleading, however. Scholars will continueto attempt to appropriate the term “leadership”for their own particular purposes, maintaininga sterile debate over the meaning of the word.However, a reading of the research reveals fivenecessary conditions for leadership on whichall the authors reviewed here appear to agree,though usually implicitly. Leadership is rela-tional, asymmetric, salient, domain specific, andinstrumental. For the kinds of leaders that con-cern us here, it also has two other characteris-tics: It occurs within durable organizations, andthe leader has enforcement as well as persuasivepowers.

Leadership is relational; one cannot be aleader without followers. The dyadic (or mul-tilateral) nature of leadership implies that wecannot understand the role of the leader unlesswe also have some ideas about the objectivesand resources of the (potential) followers.

The leader-follower relationship is asym-metric. The leader enjoys attention, andpossibly loyalty and obedience, from the mem-bership, but the relationship need not work theother way around. Even if we consider a leaderto be the “agent” of the membership (Fiorina& Shepsle 1989), the leader still possesses aposition distinct from those on whose behalfshe is acting. She may be the hub for informa-tion flows, the person to whom people turn fordirection, the one able to coordinate action, orall of the above. This asymmetry is reinforcedif the leader also has the power to compel.

One form of this asymmetry can be mademore specific: A leader is salient—that is, a

leader commands her followers’ attention. Butsalience means more than that. Part of the at-traction of a leader is that the followers believethat other followers are paying attention to thesame leader. Thus, generalized salience or com-mon knowledge (see Chwe 2001) is a necessarycondition for being a leader of some population.

Leadership is domain specific. An individualwho commands the attention and respect offollowers on certain matters may be a followerin others. We can think of a CEO in relation tothe principal at her children’s school. The CEOmay be the leader of a firm, but it is the princi-pal who has the power to direct and coordinatethe children’s education as long as they attendthat school; the principal also has the last wordon who advances and who is expelled. Thismay draw too bright a line, however. As we willsee below, high-salience individuals may findways to expand the domains over which theyattract attention. Prospective followers mayattribute generalized knowledge or skill to aparticular leader extending outside the domainof initial relevance. Empirically, experimentalsubjects appear to attach status to individualsbased on their superior performance on tasksirrelevant to the question at hand.

Fifth, a leader is instrumental—she is at-tempting to get the followers to do something.This common notion of instrumentality canobscure important differences in how authorsconceive of leadership. There are at leastthree different ways leaders act instrumentally.The first, and most common, builds on thefoundation of coordination dilemmas. Agentsare assumed to have the same objectives, knowthe structure of the game, and understandhow their actions map directly onto theirindividual and collective welfare. They justneed some help acting in concert. To theextent there are multiple equilibria, leaders cantilt the group decision toward their preferredoutcome. A second way in which leaders can actinstrumentally occurs when potential followersmay not fully understand the strategic situationin which they are embedded. For example,certain people may not even realize that it is intheir interests to coordinate. They may fail to



recognize that there is a common purpose oragenda. Here the leader does even more: Sheprovides a model of the world that enables fol-lowers to then impose some structure on theirsituation. Third, a leader may possess specialcharacteristics—oratorical skill, a genuinelynew idea, or perhaps “charisma”—such thatthe leader is able to transform the beliefs andpossibly the preferences of the followers.

Asymmetric relationships, salience, domainspecificity, and instrumentality—jointly orindividually—capture only part of what is ofinterest about political and economic lead-ership. Pop stars and sports icons can havethese features, but few are considered leaders,notwithstanding parental handwringing aboutthe quality of role models. We note two addi-tional important considerations in developing apolitical economy of leadership: Leaders oper-ate within durable organizations and, therefore,possess a capacity to enforce. Our interest is inthe discernable effects of leaders on observablegroup behavior, particularly within the contextof defined organizations (as opposed to a loosecollection of people at a ball game or in acity plaza). We restrict our examination ofleadership to the context of relatively durableorganizational relationships and excludemoments of ephemeral group instigation.

This emphasis on individuals who operatewithin the context of an organization has ad-ditional implications. Once a salient individualmoves from being simply salient to occupyinga specific office, that person commands morethan attention. The individual also gains coer-cive powers to a greater or lesser degree. High-level managers can dismiss workers. Religiousleaders can condemn the actions of followers,possibly excommunicate them, and certainlyhave them ostracized. Leaders in some orga-nizations have recourse to physical force. Aswe will emphasize below, the processes throughwhich these organizational leaders try to garnerconsent and effort from the membership—andto what end—necessarily occur with some de-gree of coercion in the background. In the lan-guage of Henrich & Gil-White (2001), leaderscan be “forceful” or “persuasive.” Leaders have

a repertoire of tactics and resources on whichto draw, but thus far, theory and empirics tendto shy away from discussing coercion as one ofthem.1

STRUCTURAL ANDINFORMATIONAL THEORIESOF LEADERSHIP

The political economy of leadership hasemerged from two distinct modeling traditions:the spatial voting model, with its corresponding“chaos” theorems, and the economics of infor-mation, particularly models of the firm. In orderto better organize current theoretical work, werefer to the former as structural theories of lead-ership and the latter as informational theoriesof leadership. In structural theories, the leaderis someone who occupies a particularly impor-tant position in some predefined institutionalstructure. By virtue of her position, the leaderis by definition salient. She can then use her po-sition to influence the organization in ways sheprefers. In informational theories of leadership,the pressing problem is typically one of coor-dination. The leader provides a common focalpoint around which the group may coordinate.

It is worth noting that this distinction be-tween information and structure, while usefulfor our purposes, breaks down rapidly oncetaken to the real world. Leaders generally oc-cupy offices in predefined organizational struc-tures and acquire and disseminate informationby virtue of holding that office.

Structural Theories of Leadership

The role of leaders in the social choice tradi-tion derives from the well-known instabilityin social choice (Arrow 1951) and “chaos”theorems of voting games more specifically.

1Chwe (1990) is one exception. Although theorists of the statenecessarily consider coercion, much of the political economicliterature concerns itself with bargaining among well-definedparties, neglecting those who are outside the bargaining pro-cess and are simply compelled to go along with the deals elitesmake among themselves (Moe 2005).

6 Ahlquist · Levi


McKelvey (1976, 1979) shows that when policyis multidimensional, you can reach (almost)any point in the policy space through some se-quence of votes. We also know that importantinstitutional arrangements, specifically agendacontrol, can induce stable equilibrium votingoutcomes (Shepsle 1979, Shepsle & Weingast1981). Taken together, these results seem toimply that if you can control the agenda, youcan get from any status quo to (almost) anydesired outcome. Leaders who know how tomanipulate the institutional situation can moldoutcomes to their own benefit or can tweakinstitutional design to suit their goals (Arrow1951, Plott & Levine 1978).

Riker (Austen-Smith & Riker 1987, 1989;Riker 1982, 1984, 1986, 1996) builds on theseresults in developing his notion of “heresthet-ics.” Rikerian leaders occupy a specific office ina predefined structure, enabling them to ma-nipulate the levers of choice (e.g., agenda con-trol). But more interestingly, some leaders canalso affect the dimensionality of the policy spaceand in so doing radically alter the outcomes.

Two recent, important works build directlyon this foundation, albeit in quite differentways. Canes-Wrone’s work on the Americanpresidency (Canes-Wrone 2006, Canes-Wroneet al. 2001) can be understood through the lensof Rikerian heresthetics. She seeks to explainwhy and when presidents take their policy ap-peals directly to mass publics and when a theyadvocate policies not immediately popular withthe electorate. She first models the strategiclogic of public appeals and then turns her at-tention to presidential “pandering,” defined asa president’s advocacy of a popular policy evenwhen he believes the policy is not in the coun-try’s best interests.

For Canes-Wrone and coauthors, presiden-tial leadership means pursuing a policy the pres-ident believes to be in the country’s best inter-ests despite a dearth of popular support at thetime the policy is chosen. Voters can come tounderstand the wisdom of such leadership, butthis is not certain (Stokes 2001a,b). To modelthe presidential choice, Canes-Wrone proposesa game between the president, a challenger, and

voters. There is a policy issue about which thepresident has private information. Politiciansalso vary in their competence or inclination toimplement a particular policy. Thus, there aretwo points of asymmetric information betweenvoters and the president: the president’s com-petence and the state of the world. After re-ceiving his information, the president choosesto exercise policy leadership if he implementsthe policy; by failing to implement, the presi-dent panders to public opinion. If the presidentexercises policy leadership, voters either learnthe policy outcome or become aware that theywill not know the outcome until after the elec-tion. Voters then choose the incumbent or chal-lenger and the game—save for the election—isrepeated for the second term.

The model predicts that a president willexercise policy leadership when (a) the presi-dent is highly popular, highly unpopular, or inhis second term; and (b) the president believesthe public misapprehends the optimal policy.A president has incentives to pander to publicopinion when he is running for election, he ismarginally popular, and policy consequenceswill not be known for some time. Accordingto Canes-Wrone (2006), any variation in thecongruence between the policy positions ofthe president and those of the electorate mustbe due to pandering. She provides both casestudy and statistical evidence consistent withthe model’s predictions.

Her work has several important im-plications. First, she shows that even if aparticular individual has arisen to a positionof high salience—in her application, the U.S.presidency—the way in which the presidentchooses to deploy informational resourcesfollows a predictable strategic logic. Under-standing the office and the strategic situation isperhaps more important than the occupantof the office at a particular moment (alsosee Saiegh 2009). Maybe our world is notso chaotic after all. Second, Canes-Wroneshows sensitivity to the empirical challenge ofdisentangling the office from the individual.Specifically, to test the implications aboutpandering and leadership, she needs to present



convincing evidence that the president hasbeliefs about the appropriate policy. Thisalone requires detailed historical knowledge.But to make the case more persuasively, sheshows instances of the same individual initiallytaking actions opposite public opinion but thensubsequently pandering.

Schofield (2006), though working out of asimilar tradition, takes a radically different tackin exploring “architects of change” at criticalmoments in time. He argues that any set ofinstitutional arrangements relies on “core be-liefs.” These beliefs take on a number of inter-pretations, but a critical property is that some-how the relevant population has come to holdthem in preponderance. What’s more, core be-liefs constitute a commonly shared model ofhow others formulate their beliefs, preferences,and, ultimately, choices. In a world of uncer-tainty, core beliefs are Schofield’s glue holding adynamical system together. For Schofield, piv-otal moments in history—what he calls “consti-tutional quandaries”—occur when core beliefsno longer conform to reality or, slightly moreformally, when there is extreme variation in in-dividual assessments of the appropriate actionrelative to the common, shared understanding.When these moments occur, small shifts in thebeliefs of a few may trigger a cascade that re-sults in an entirely different configuration ofbeliefs.

A leader is the agent most likely to triggersuch a cascade. According to Schofield, “archi-tects of change” must do two things: (a) com-municate a model of the world in which thereare specific outcomes associated with differingcourses of action and (b) convincingly advocatea specific outcome. We can make an analogyto Hans Christian Andersen’s story of the em-peror’s new clothes: Although the child whoidentified the emperor’s nudity precipitated a“quandary,” the child himself was not an archi-tect of change. An architect of change must alsoarticulate a story of what the emperor’s nuditymeans and what should be done as a result.

Schofield gives several examples of “archi-tects” from the past 250 years, but he readilyacknowledges that where these architects come

from, how they garner the attention of largenumbers of people, what makes their argumentsmore compelling than others, and how “beliefcascades” function are all outside the scope ofhis analysis. Majumdar & Mukand (2010), how-ever, show that marginal changes in a leader’sjudgment can have discontinuous effects on theprobability of successful group mobilization.

Given their common inspiration in thesocial choice and spatial modeling traditions,the Canes-Wrone and Schofield conceptionsof leadership could not be more different. ForCanes-Wrone, the leader is an office holderseeking to navigate a well-defined policyspace in the context of robust and exogenousinstitutions. Schofield discusses leadership inthe absence of virtually all these elements, saveperhaps the defined policy space. From theperspective of other social scientific works onleadership, Canes-Wrone and Schofield areexamining different types of leadership (Burns1978, Hargrove 1989). Canes-Wrone’s projectexplores the day-to-day functioning of the U.S.government and considers the president as a“transactional” (Hargrove 1989) leader whomanages and attempts to implement policyin the context of a larger political and socialequilibrium, an equilibrium that includes thedimensions of the policy space and the beliefsheld by legislators and voters. Leaders andfollowers all know the game; the leader simplyhelps manage the process of dealing withuncertainty. Schofield, on the other hand, isinterested in “transformative” (Hargrove 1989)leaders who induce discontinuous change atpivotal moments. His architects of changeredefine the game that the followers believethemselves to be playing.

Informational Theories of Leadership

A common political economic approach to lead-ership is in the context of a pure coordinationgame in which players have symmetric prefer-ences but need help coordinating on the sameaction or strategy. In such a set up, agentscan use commonly observed cheap-talk signalsto coordinate on one equilibrium or another

8 Ahlquist · Levi


(Frohlich et al. 1971). Relabeling this signal as“leader” then follows naturally. From this sim-ple beginning, several intriguing models haveemerged in recent years. Some consider thestructure of the game and payoffs to be com-mon knowledge, focusing instead on credibil-ity. Others assume fully aligned preferences andexamine the role of leaders in affecting the di-rection of an organization and even the endoge-nous emergence of leaders.

One of the earlier models explicitly allow-ing for both a leader and learning is by Bala &Goyal (1998). They model a large-scale coor-dination problem in which agents are uncertainabout which of two actions is optimal. In for-mulating their beliefs, agents learn from theirimmediate neighbors but also observe a “royalfamily.” These widely observed agents turn outto have a strong influence on equilibrium selec-tion simply by virtue of their commonly knownsalience.

Other work emerges from the contractar-ian tradition (Coase 1937), building on a spe-cialized version of the collective action/publicgoods problem generally referred to as “moralhazard in teams.” These models derive fromprincipal–agent questions focusing on how toelicit optimal levels of effort (from the princi-pal’s perspective) from agents under differentinformation structures, levels of risk aversion,and more or less complicated organizationalstructures.

In the most basic set up, there is some well-defined group of individuals. Each individualcan contribute costly effort to the group project,deciding noncooperatively how much effort toexpend. Each individual’s effort is unobservableand unverifiable. It is also generally assumedthat agents cannot be coerced into supplyingeffort. There is some function that turns cumu-lative effort into output, but productivity variesstochastically. Although team output is observ-able and can be contracted upon, the agentsmust choose their level of effort before theylearn the value of the random productivity com-ponent. It is effectively impossible to write anenforceable contract stipulating each person’seffort level.

If, as is generally assumed, the total outputmust be completely divided up, i.e., the sumof the payments to the team members cannotbe more or less than total output, there is theadditional condition referred to as the budget-balancing constraint. Holmstrom (1982) showsthat, where there is a team production problem,there is no Nash equilibrium incentive contractthat is both budget balancing and Pareto effi-cient. The intuition behind this result is thatindividuals will only contribute to the groupproject up to the point at which their individ-ual marginal gain is equal to their individualmarginal cost, even if their marginal cost of ef-fort is far lower than the marginal team-leveloutput.

Hermalin (1998, 2007) and Komai et al.(2007, Komai & Stegeman 2010) look at therole of a leader in the context of moral hazardin teams. Incomplete information is the funda-mental rationale for leadership, but it is not im-mediately clear that it is sufficient. If all the teammembers have exactly the same objective func-tion and access to private information of similarquality, then, according to the “Jury Theorem”(Condorcet 1785), simply taking votes shouldyield the optimal decision. But a simple vot-ing solution misses two important pieces of thepuzzle: uncertainty about the beliefs of othersand the possibility of some competing interests(i.e., a strategic situation). Hermalin incorpo-rates both these issues by modeling leaders asagents with privileged and valuable informationand the undivided attention of the other teammembers. The leader is salient and this salienceis common knowledge.

Leaders in Hermalin’s world know thevalue of the group’s productivity parame-ter. Although this information is valuableto everyone, a problem arises because theleader has an incentive to lie: The leaderalways wants to claim that the group is inthe “high-productivity” state to induce all theother team members to put in their maximumeffort, even if she puts in less. The otherteam members, anticipating this, disregardthe leader’s statement. Hermalin considerstwo possible ways the leader can credibly



transmit her valuable knowledge: “sacrifice”and “leadership by example.” In the first, theleader makes a gift to the members out of herpocket. In the second, the leader exerts (costly)effort in order to influence others to follow suit.

The sacrifice solution is really just a mecha-nism design problem. Is there a contract thatinduces the leader to tell the truth and thatmakes the leader better off than in the situa-tion in which the team fails to use her knowl-edge? Hermalin proves that there is; in thehigh-productivity state, the leader distributessome fixed amount to the other team members.This amount is just big enough that a leader inthe low-productivity state has no incentive tocopy. Note that this solution relies on the en-forceability of this contract and only allows theteam to take advantage of the leader’s superiorknowledge. It does not overcome the team pro-duction problem. The other team members stillonly contribute their individually rational lev-els of effort. The sacrifice option is still secondbest.

In the leadership-by-example solution,Hermalin relaxes two assumptions in the stan-dard team production game, one explicitly andone implicitly. First, he explicitly allows theleader to expend her effort before the rest ofthe team decides how much effort to contribute.Second, he implicitly assumes that team mem-bers’ effort levels are observable but unverifi-able, i.e., all other members of the team canobserve the leader’s effort at no cost, but stillno enforceable contract can be written on thebasis of effort levels. In leading by example, theleader can signal to the team whether they are ina high- or low-productivity state simply by hereffort levels. Because this action is costly, it iscredible. The lead-by-example outcome is su-perior to the sacrifice option, since the leader’saction directly contributes resources to theteam project, increasing the surplus to be sharedby all in addition to inducing effort from thefollowers.

Hermalin (2007) extends his initial modelto allow for repeat interactions between the(infinitely-lived) leader and (short-lived) mem-bers, thus allowing leaders to build reputations.

The value of such a reputation is in its abilityto economize on signaling costs. The more fre-quently a leader would otherwise have to signal,especially using the sacrifice strategy, the morevaluable the reputation.

Recent extensions of Hermalin’s model(Komai et al. 2007, Komai & Stegeman 2010)consider situations in which the choice of ef-fort is binary (participate or not) as opposed tocontinuous. In this restricted action space, theleader’s choice of effort now fails to fully revealthe state of the world; followers can be inducedto contribute in situations where they would notif they knew the state perfectly. “Because therecan be a benefit to keeping information from thefollowers, a group may have a motive to createa leadership position and limit how much infor-mation the leader can credibly communicate tothe group” (Komai et al. 2007, p. 947, emphasisin original).

But even with a Hermalin-style leader inplace, outcomes still fail to achieve first-bestlevels of effort. Miller (1992) summarizes theimplications of a long line of research into theoptimal incentive contract with the dictum: “In-formation asymmetries . . . and production ex-ternalities make it impossible for managers torealize the full efficiency potential of team pro-duction processes through the manipulation ofshort-term economic incentives alone” (p. 198).

These firm-oriented models are about op-timal ways to divide up a surplus to elicit ef-fort and compliance. Mapping theories of firm-based production onto other organizations suchas unions, political parties, or churches requiresreconsideration of the nature of the “good” pro-duced. Unlike the continuously divisible out-put assumed in models of firms, the fundamen-tal “good” produced by voluntary organizationsis a club good, i.e., it is nonrival. The polit-ical levers in different organizations availableto leaders vary, often in accordance with thegroup’s objectives. For example, labor unionscollectively bargain for an employment con-tract; the benefits of a union contract cannot beallocated across individual members to createany system of incentives for member contribu-tion to the union.

10 Ahlquist · Levi


Furthermore, the relationship betweenrank-and-file union members and the unionleaders differs from the manager-employee re-lationship. It is more similar to that betweenlegislators and chief executives as analyzed byFiorina & Shepsle (1989), who emphasize theconundrum of whether it is better to considera leader as a principal or as the agent of hernominal followers. The union leader cannot al-locate wages directly to union members con-ditional on the quality of the union contract.Rather the leader is, in some ways, the agent ofthe membership and must convince members tocontribute through dues and, when necessary,picketing and striking. But once the union’s or-ganizational rules are in place, the leader willpossess the power to enforce them. Althoughshe may not easily exclude or expel someonefrom the union, as the holder of the leadershipoffice, she will nonetheless have other means ofenforcement, ranging from social pressure andpublic humiliation to fines.

Additionally, many scholars of the firm pre-suppose that the task, production function, andobjectives of all the team members are well-defined and common knowledge. The Herma-lin model and extensions are silent on howthe leader and followers coordinate on whatthe organizational objectives are in the firstplace, leaving little room for the inspirationalrole leaders seem to play in the real world,where organizational goals, tactics, and the be-liefs of others are frequently uncertain or evencontested.

Dewan & Myatt (2007) offer a model morerelevant to politics: There are common inter-ests among group members and the leader butuncertainty about both the exogenous state ofthe world and the beliefs of others. The funda-mental problem is coordinating a large groupon one of two actions without the optimal ac-tion being known. The Dewan & Myatt modelalso makes an additional Michelian assumption:The followers, “activists” in this vocabulary,want to be told what to do.

In their model, each activist receives a noisyprivate signal as to the preferable course ofaction, and the precision of this signal con-

stitutes that individual’s “sense of direction.”Activists evaluate the information they receiveand, in a manner similar to that in Feddersen &Pesendorfer (1998), calculate the relative like-lihood of being pivotal for one course of action.Ultimately, each activist decides which courseof action to advocate. If a sufficient number sup-ports a particular policy, the entire group getsthe payoff associated with that action. If thegroup fails to generate a critical mass for ei-ther option, the group fails at its task. Dewan &Myatt formulate a novel solution concept akinto Bayesian perfection and show that, withouta leader, the equilibrium threshold is biasedtoward the option with higher coordinationbarriers.

Dewan & Myatt (2007) then introduce thepossibility of a leader. Similar to Hermalin,“[i]n our world, the emergence of a leadercorresponds to the introduction of public in-formation” (p. 829). More concretely, a leaderoffers up a “speech” in which she attempts tocommunicate the information she possesses.They analyze the conditions under which aleader wants to dictate a specific action, whichremoves the risk of coordination failure; oradvocate for some threshold that would bias thegroup threshold in a particular direction buttakes advantage of the information distributedacross the group. They then consider theconditions under which the members will“follow the leader.” Dewan & Myatt derivewhat they label the “Michels’ ratio,” in whichthe “need for direction” parameter governs theextent to which it matters which option thegroup coordinates on:

R ≡barriers to coordination × sense of direction

need for direction.

They show that leaders will prefer to dictatewhenever R > 1 and will advocate a particu-lar threshold whenever R < 1. Similarly, whenR < 1, group members will all adopt the com-mon threshold that the leader advocated. Thisis so because when “she announces her signal,other activists know what she knows. As theyall have the same information at the leader’s



disposal and share her preferences, they act inthe way she would like them to” (Dewan &Myatt 2007, p. 837). When R > 1, the leaderwill advocate for a particular course of action,and followers will ignore their private informa-tion to adopt her recommendation.

The Dewan-Myatt model formalizes sev-eral of Michels’ insights while providing(a) a technical vocabulary for discussing theconcept of leadership in a group and (b) a wayto understand when a leader (or leadershipcadre/oligarchy) will emerge.

Dickson (2011) develops another cheap-talkmodel in which there are competing interestsbetween the leader and the two followers, a co-ordination problem, and uncertainty on the partof the followers about the state of the worldand the interests of the leader. The leader,again, possesses valuable information and holdsa salient position; the identity of the leader istaken as given. In the model, there are threestates of the world. In two of the states (“agree-ment states”) the leader and both followers havepreferences that are aligned perfectly. In thethird state (“disagreement state”), the follow-ers are involved in a battle-of-the-sexes gamein which both want to coordinate on the sameaction but differ as to which action they wantto coordinate on. The leader has a preferencefor one of the two actions as well, but thispreference is unknown to the followers. Theleader sends a signal as to the state of the worldand then the two followers pick their actions.In the equilibrium considered, a self-interestedleader will never announce the disagreementstate, and rational followers will update beliefsabout the state based on priors about the lead-ership type and the probability that a particu-lar state obtains. Followers will coordinate onthe leader’s announcement in every state of theworld, allowing the leader to get her most pre-ferred outcome in the disagreement state. Whatmakes this situation particularly interesting isthat the leader can affect followers’ behavior toher benefit even if the followers do not believethe leader’s announcement to be credible as astatement of fact. In the language of Dewan &Myatt, the need for coordination coupled with

uncertainty can provide a salient leader with thespace to manipulate outcomes.

Not surprisingly, the critical component un-derlying these models is the transmission of in-formation within the group. The leader offers asignal, which members must be able to observeat low cost. The importance of informationfor group production harkens back to Barnard(1938) and is congruent with Arrow’s (1974)argument that the manner of information ag-gregation uniquely defines an organization. In-formation transmission affects the formationand modification of individual beliefs and thewillingness of individuals to comply with lead-ers’ demands (Arrow 1974, Levi 1997). Effec-tive communication between the leader and therank-and-file about the both the leader’s prefer-ences and information is critical for persuasion(Lupia & McCubbins 1998).

Bolton et al. (2008), in examining the or-ganizational challenge of effectively coordinat-ing group action when adapting to changingconditions,2 reveal conditions in which infor-mation flows from followers to leader as wellas vice versa. In their model, information aboutthe state of the world trickles in gradually in theform of private signals to the (exogenously iden-tified) leader and set of followers. Specifically,the leader relies on her prior beliefs to articulatea “mission statement,” i.e., her best guess as tothe state. Next the followers observe their sig-nals and take actions. The leader then observesmore information about the state and takes heraction that determines the payoffs for all. Notethat a leader in the Bolton et al. model is morethan simply a salient source of information—the leader here takes a binding action thatdetermines the outcome for the entire group.

A tension arises because followers must de-cide how much to incorporate their private sig-nals into their choice of actions, which in turndepend on how they expect the leader to re-act to new information. If the leader is not

2Dessein & Santos (2006) and Ferreira & Rezende (2007)address similar problems but without incorporating a rolefor leadership.

12 Ahlquist · Levi


expected to deviate from her announced mis-sion, then followers are better off ignoring theirown information. Bolton et al. build on pre-vious work (Blanes i Vidal & Moeller 2007;Gervais & Goldstein 2007; Van den Steen 2004,2005) by allowing the leader to be “resolute,”i.e., she overestimates the precision of her priorbeliefs and is therefore slow to change her mindas more information becomes available. Theyshow that the coordination benefits emergingfrom a resolute leader (relative to a fully ra-tional leader) generally outweigh the risk oforganizational maladaptation. The benefit de-rives from the assumption that followers knowthe true variances of all signals. The implica-tion is that followers follow even when theyknow that the leader is overconfident in her ownjudgment.

The Bolton et al. model also opens upa role for something like corporate culture(Hermalin 2001, Kreps 1990, Levi 2005, Miller1992), that is, a statement of principles delin-eating the manner in which decisions will betaken and the rights of the membership andleadership vis-a-vis one another. Suppose theleader’s signal is simply an aggregation of theactions taken by the followers. If the leader isvery resolute, the followers know any informa-tion conveyed by their actions will be ignoredand that they are better off following the mis-sion statement. A less resolute leader allows forweaker coordination but learns from the actionsof the followers, enabling better organizationaladaptation to changing circumstances. Both areequilibria. If a prospective follower is unsurewhich equilibrium the leader (and other follow-ers) have selected or cannot reasonably predicthow his organization will respond to changingevents, he is unlikely to be willing to join orcommit resources to the organization.

Endogenous Leadership

The models discussed thus far take the leaderto be exogenously identified. A few recentpapers have begun to consider endogenouslyemerging leaders. One set of papers (Huck& Rey-Biel 2006, Kobayashi & Suehiro

2005, Komai & Stegeman 2010) extends theHermalin (1998) model of leadership in theteam production problem. Although the detailsdiffer across models, each essentially showsthat, under certain conditions, leadership iswelfare improving (or even first best) and thata leader–follower situation can constitute anequilibrium. Nevertheless, the political processby which this “optimal” individual is identifiedis nowhere specified; these papers make ref-erences to leaders either being “appointed” bysome unidentified appointer-in-chief (Komai& Stegeman 2010) or they simply “emerge”probabilistically when one player or anotherdecides to move first (Huck & Rey-Biel 2006,Kobayashi & Suehiro 2005).

Dewan & Myatt (2008, 2009) take a morenuanced step toward endogenizing the emer-gence of a leader (also see Majumdar & Mukand2010). Again the fundamental social dilemmathey build upon is a coordination problem.Agents advocate a particular policy in a one-dimensional policy space. They care about theappropriateness of the policy for the (imper-fectly known) situation as well as unity withothers in the group. A key assumption in themodel is that agents are ex ante symmetric andthat this property is common knowledge amongall agents.

Leaders are again modeled as “informationsources” (Dewan & Myatt 2008, p. 356) or “in-formative signals” (p. 354) whose preferencesare aligned perfectly with those of the mem-bership. Potential leaders observe private, in-dependent signals as to the state of the world;the signal is correct in expectation but may beimperfectly received by the leader. The preci-sion of a leader’s understanding is her “senseof direction.” Each (potential) leader can thenaddress the membership by sending a message.The message the leader sends to the member-ship, however, is noisy; its precision is referredto as the leader’s “clarity of communication.”

Agents must decide how to map the mes-sages they receive (i.e., including any distortiondue to a leader’s misapprehension or miscom-munication) into an advocated policy that takesinto account both the agent’s updated beliefs



about the state of the world and the updated be-liefs about others’ beliefs. In the unique perfectBayesian equilibrium, the weight a follower as-signs to a leader is increasing in both the leader’ssense of direction and her clarity of communi-cation. As the group’s concern for cohesion in-creases, the clarity of communication becomesmore important than correct apprehension ofthe environment.

Dewan & Myatt’s key advance is extendingthe model to allow agents to endogenouslyallocate attention to particular leaders. Fixingthe clarity of the leader’s message, the agent’sunderstanding of that message improves ashe devotes more attention to that leader. Inequilibrium, agents only listen to the clearestcommunicators, and among that elite groupof clear communicators, those with a strongersense of direction receive proportionately moreattention. Furthermore, if the clearest com-municator is sufficiently better than the secondbest, she can become a de facto “dictator” inthe sense of enjoying exclusive attention fromthe membership.3

The equilibrium in the strategic clarity gamehas a couple of interesting comparative staticproperties. First, an increase in any leader’ssense of direction weakly increases the clarityof communication of all leaders. The intuitionbehind this result stems from the idea that anincrease in a leader’s sense of direction forcesthe others to compete harder to retain the at-tention of the members. Second, as the weightthat agents place on group unity relative topolicy increases, both the clarity of communi-cation and the average appropriateness of theagents’ advocated policy increase. When mem-bers “desire unity, they seek out a commonparty line and so leaders respond by speakingclearly. Clearer communication allows activiststo develop a better understanding of their en-vironment” by shifting attention to the leaders

3If the best leaders are so talented that they can get their pointacross quickly they may even have an incentive to “obfuscate,”i.e., speak less clearly than they are capable of doing so as tomaximize the attention they receive from the followers.

with the strongest sense of direction (Dewan &Myatt 2008, p. 363).

Although Dewan & Myatt make an im-portant advance in allowing “followers” toendogenously choose their leaders, they also ex-pose important gaps. First, in the Dewan-Myattworld, the set of candidates for leadership po-sitions is exogenous. An obvious area for futuremodeling work is allowing individuals to investin acquiring information, oratorical skills, orboth and then letting them declare their desireto “lead.” Second, the attention-maximizationmotive surely drives some (potential) leaders,but it seems reasonable to assume that mostindividuals who desire the attention of a groupdesire it for some end that may or may not coin-cide with what the membership wants ex ante,an issue Dewan & Mayatt only begin to address.

Returning to our economics-of-informationframework, these recent modeling exercisesexplore the two important issues of competinginterests and coordination, but at the cost oftreating them separately. Hermalin (1998,2001, 2007) uses uncertainty about the stateof the world to generate a role for a leaderand explores the conflict of interest betweenthe leader and the group, but he ignores co-ordination problems. Bolton et al. (2008) andDewan & Myatt (2007, 2008, 2009) introduceuncertainty about both the exogenous stateof the world and the beliefs of others, butthey consider only common interests. Dickson(2006, 2011) incorporates some measure ofdivergent preferences and coordination butrequires the leader to be fixed and exogenous.No model so far encapsulates noninformationaltools available to leaders, such as coercion,excommunication, and demotion; nor is theresufficient attention to competing interests,coordination, and endogenous leadership.

Leadership and the Formationof Beliefs and Preferences

How followers develop expectations about thelikely future behavior of both the leader andtheir fellow members lies at the heart of boththe Dewan & Myatt (2007, 2008) and Schofield

14 Ahlquist · Levi


(2006) pieces. The Dewan-Myatt world, popu-lated by the standard game-theoretic Bayesianmaximizers, lacks a role for leaders in affectingthe scope of the organization and some con-ception of how leaders might act to inspire andtransform members’ behavior. Schofield takesa more complicated and less well-defined tackto identifying core beliefs and belief cascades.Although he presents a promising schematic forunderstanding discontinuous change (at leastex post), he still nominally retains the assump-tion of Bayesian agents while readily admittingthat he has no way of explaining where a leadercomes from, what makes one more convincingthan another, or how information transmissionmay or may not result in a belief cascade.

Dickson (2006) makes an importantcontribution here by explicitly relaxing theassumption of Bayesian agents. He showsthat, from an evolutionary perspective, non-Bayesian learning can be evolutionarily stablein a coordination game not dissimilar to thestrategic situation described by Dewan &Myatt. Simulation work (Siegel 2009) andmodels of network interaction and networkgames (Chwe 1999, 2000; Jackson 2008) haveshown that the structure of network connec-tions among agents has strong effects on thedistribution of beliefs and, ultimately, choicesacross that group, providing some foundationfor modeling how leaders can influence othersand whether and how belief cascades mightoccur. The fully connected yet locally inde-pendent network structure of Bala & Goyal(1998) is but one of many possible structures.This work reinforces the importance of contextand the details of the situation in order to linkgeneral theories with organizational reality.

Exploring how organizations and leaders af-fect the members’ beliefs and behavior is apromising and well-defined avenue for research(Levi 2005). As Dickson (2006, pp. 10–11)points out, “the best way to learn (or update be-liefs) depends on what it is exactly that othersdo—and how it is exactly that others are learn-ing” and “group membership can affect individ-ual cognition in equilibrium because individualbelief updating behavior is itself determined in

equilibrium” (p. 11). Dickson’s argument com-ports with Henrich & Gil-White’s (2001) no-tion of prestige, in which agents willingly deferto select others.

EMPIRICAL WORK

Much of the recent wave of research on leader-ship has been theoretical. Nevertheless, empir-ical regularities, though harder to come by, arebeginning to emerge.

A basic question is whether there is any evi-dence that the identity of the leader makes a dis-cernable difference for policy or organizationalperformance. The key methodological problemis breaking the obvious endogeneity betweenwho gets to lead and the leader’s performance;elected leaders who perform poorly are likelyto be voted out. A few studies have begun theprocess of resolving this thorny issue, but thereis still considerable work to be done.

Jones & Olken (2005) exploit instances inwhich heads of state die in office from naturaland accidental causes to examine the effect ofexogenous leadership changes on rates of eco-nomic growth and other policy outcomes. Evenwith only 57 cases and a dependent variable ascomplex as economic growth, they find consis-tent evidence that changes in growth rates areassociated with changes in the identity of thenational leader. Furthermore, these effects aremost pronounced in autocracies. We can in-terpret the Jones & Olken findings as evidencefor an important structural role for leaders andleadership: the individual occupying a positionof peak power and authority in government canaffect macro-level outcomes independent of theforces that installed that individual in the firstplace.

Humphreys et al. (2006) look at a moremicro-level setting with findings relevant forboth the structural and informational theoriesof leadership. Their solution to the method-ological problem of endogenous leadership isa field experiment in which they randomly as-signed discussion leaders to a series of commu-nity meetings from which the government ofSao Tome and Principe sought to infer some



sort of consensus on how to approach the coun-try’s recently discovered oil wealth. Humphreyset al. seek to determine whether there is a “lead-ership effect” in a deliberative democratic set-ting, i.e., does the identity of the discussionleader affect the group’s expressed preferences?Group leaders presented information and thenled discussion on a variety of preselected top-ics. The leader then recorded the group’s pref-erences on a series of questions as well as thedegree of consensus achieved. Using a leaderfixed effects model, Humphreys et al. find thatbetween one-fifth and one-third of the cross-group variation in expressed preferences can beattributed to the identity of the leader. Whenconsidering consensus, “two-thirds of the varia-tion in the reported level of consensus achievedduring the discussions can be accounted for onthe basis of leader-specific fixed effects alone”(Humphreys et al. 2006, p. 608). They presentsuggestive evidence that the leaders shifted theexpressed group preferences toward the out-comes the leaders preferred.

Because the groups made decisions throughconsensus, appeals to heresthetics or evensimple agenda control cannot explain the out-come. Furthermore, it appears that leader char-acteristics such as gender and age had effectsthat are not obviously subject to immediatemanipulation by the leader herself (also, seeChattopadhyay & Duflo 2004).

The Humphreys et al. findings seem to posea challenge for the economics-of-informationschool. They offer no evidence of learning,persuasion, or leadership communicationof information. Humphreys et al. concludethat participant self-censorship or leadermisreporting are the most likely explanationsfor the consistent leadership effects they find.However, further study in different settingsand with different treatments is surely neededto illuminate the mechanisms at work and toassess the robustness of the findings.

A second body of empirical research drawsdirectly on informational theory. It focuses noton who holds the position of leader but onwhether there is a leader and the mechanismsby which a leader can improve group out-

comes. The simple coordination game posesthe question of whether a salient individual canhelp groups overcome coordination problems.It turns out the answer is “yes,” with some in-teresting subsidiary findings.

Wilson & Rhodes (1997) constructed anexperiment in which groups played a simplecoordination game and one player at randomwas selected to be the leader. The leader sent aprivate signal to each follower. Put in the termsof the discussion thus far, the leader was salientand the salience was common knowledge,but the signals could vary across followers.Wilson & Rhodes found that having a leaderdoes increase coordination significantly butfails to result in full coordination. They alsomanipulated the extent to which the leader’sand followers’ interests were aligned. Onlywhen the “good” leader was very likely didfollowers obey the leader’s recommendation.Their findings “indicate that uncertainty abouta leader’s type can markedly affect the impactof leadership. Subjects are neither capableof making fine distinctions across levels ofprobability nor do they precisely recognizethe thresholds at which leader signals remaincredible” (Wilson & Rhodes 1997, p. 787).

Eckel & Wilson (2000, 2007, 2009) con-ducted a series of experimental coordinationand public goods games in which the “leader”(again simply a player observed as salient byall the other players) was either “high status”or “low status,” with status determined by theplayer’s performance on a trivia quiz. In gen-eral, higher-status individuals were mimickedmore regularly than lower-status individualsand were also more successful at moving thegroup to superior equilibria. The status ef-fect translates prowess in one area to unrelatedstrategic situations, underscoring the perme-ability of leadership’s domain specificity. How-ever, the finding has limited importance, giventhat the conditions affecting the generalizationof influence are not well understood outside theexperimental setting.

Several papers have sought to explore differ-ent predictions of the Hermalin (1998) model inan experimental setting, generally focusing on

16 Ahlquist · Levi


leadership by example. Recall that the strate-gic situation is a voluntary contribution gamewith divisible output and a stochastic produc-tivity factor. The leader is the player who knowsthe value of the productivity variable and canmove before the follower(s); the leader’s moveis observed by the follower(s).

Of those using experiments to explore theleader–follower game, all except Komai &Grossman (2009) use full-information versionsin which both the leader and follower(s) knowthe value of the state variable but the leaderstill gets to move first (and have her moveobserved by all). In the full-information game,subjects are more likely to contribute afterhaving observed the leader’s contribution thanin situations where everyone must move si-multaneously. For example, Moxnes & van derHeijden (2003) document leadership-by-example effects in sequential, full-information,public-bad games, although the effect degradesover experimental iterations. It appears leaderscan set an example even where they do not needto transmit information. A common findingacross these papers is evidence of “reciprocity”or “emulation,” with followers typically doingbetter than leaders.

The experiments pose the question ofwhether signaling or reciprocity explains howadding a leader improves cooperation. Onthis there is some dispute. Certain carefullyexecuted designs reveal a strong signaling rolefor leaders, but in more complicated and, ar-guably, more realistic settings involving repeatplay, other considerations become relevant. Inthe experimental set up of Potters et al. (2007),(a) each value of the productivity factorimplies a specific equilibrium prediction, and(b) followers and leaders are shuffled randomlybetween rounds of play. They find that inthe incomplete-information setting, leadingby example increases total contributions by50% and follower contributions by 73% com-pared to a simultaneous-move (no leadership)benchmark. Furthermore, this increase infollower contribution rates is much largerunder incomplete information than undercomplete information, implying a strong and

possibly dominant role for signaling comparedto reciprocity. On the other hand, Meidinger& Villeval (2002) find evidence for reputationand reciprocity. They use an experimentaldesign in which leaders and followers are pairedtogether for the duration of the experimentwith this fact known to the participants.

Guth et al. (2007) extend the complete-information model in two interesting direc-tions: They endow leaders with the power tomete out punishment in the form of exclusionfrom subsequent rounds of play, and they allowfor different leadership-selection mechanismsincluding rotating leadership and endogenousleadership selection. They find that groups withleaders capable of punishing followers have sig-nificantly higher contribution rates than groupswith no leader or “weak” leaders. When leadersare exogenous, it does not matter whether theleader holds her position for the entire exper-imental episode or whether leadership rotates.Once leadership is endogenous, only 40% ofgroups are able to successfully appoint a leadereven though groups with leaders perform muchbetter than those without. This finding of Guthet al. must be taken with a few grains of salt,however, because their leader-selection processrequired groups to achieve consensus on whomthey would be willing to accept as a leader with-out engaging in any communication. Intrigu-ingly, leaders who benefit from exclusion powergenerally want to remain leaders but the follow-ers often demur, whereas leaders without theability to sanction followers generally refuse totake another term as leader even though theirfollowers want them to do so. Followers appearreluctant to “take their medicine” in the formof allowing another to punish them for free rid-ing, even if this punishment is in the followers’best interests.

A recent pair of papers reveals that the ear-lier findings may not be robust in more interest-ing group settings. Komai & Grossman (2009)use a standard sequential contribution game un-der incomplete information, manipulating thesize of the group. They find that the effective-ness of leadership by example degrades signifi-cantly as the group size increases from four to



nine. Levati et al. (2007) use a set up nearlyidentical to that of Guth et al. (2007) but al-low players to have different initial endowmentsfrom which to contribute. Although groupswith leaders showed significantly higher con-tribution rates than those without, heterogene-ity reduces the influence of a leader in inducingcontributions. More interestingly, heterogene-ity eliminates any benefit from endowing theleader with exclusion power, and, in situationswhere the degree of group heterogeneity is un-known to followers, leadership has almost no ef-fect on contribution rates. It appears that whenthe leader sets an example but there is notableinequality in the group, the leader’s example isinsufficient to improve contribution rates.

Dickson’s (2011) experimental findingsalso pose a challenge for the economics-of-information perspective. Followers in hisexperiments were overly credulous of theleaders’ statements, failing to take into accountthe leaders’ incentives to misrepresent thestate of the world. Recall that in the Dicksonmodel/experiment there are three states of theworld; the leader and followers perfectly agreeon the correct course of action in two statesbut disagree in the third. In all three states, thefollowers want to coordinate on an action, butin the disagreement state, their preferencesdiverge. Leaders, too, have preferences overwhich action to take in the disagreementstate. A leader knows the true state and herpreferences perfectly whereas the followers areuncertain about both pieces of information.Dickson expects that leaders’ announcementsas to the state of the world will prove focal forthe followers as they choose actions, and in hislaboratory experiments, he finds consistent evi-dence that followers use the leader’s announce-ment as a focal point. Followers coordinate atvery high rates whenever the leader announcesan agreement state but at a lower rate whenthe leader announces a disagreement state.

Dickson, unlike the other experimenters re-viewed, makes a small bow toward external va-lidity. He solicits followers’ posterior beliefsabout the true state and turns up some inter-esting findings. He finds that, first, followers

coordinate at slightly lower rates when theleader’s announcement is not credible; thisshould not occur if agents are fully rationalBayesians. Second, followers are systematicallyfailing to account for the leader’s incentives tomisrepresent the state. Even when the leader’sannouncement of an agreement state could notbe credible given the priors for leadership typeand states of the world, followers guessed thatthe leader was telling the truth about 40% of thetime. Furthermore, Dickson finds that experi-mental subjects do not improve in this regardas they gain experience playing the game, nordo they seem to be reacting to the possibilitythat some leaders may place a value on being“honest.” He concludes that followers’ “beliefsabout the state of the world are systematicallybiased away from Bayesian judgments, towardsmessages about the world that they receive fromLeaders” (Dickson 2011, p. 29).

This initial wave of rigorous empirical workhas undoubtedly posed many interesting ques-tions that will keep researchers busy for manyyears. But two key shortcomings are worthnoting. First, the artificiality of the settingsand the relative inattention to external validitypose serious limits to what these experimentstell us about how the world actually works(McDermott 2002). Second, in all but two ofthe pieces reviewed here (Levati et al. 2007,Guth et al. 2007), the emergence of the leaderis completely exogenous. Even in these two ex-ceptions, the process for leader selection em-ploys a very particular voting mechanism thatsets a high bar. In general, the experimental ma-nipulation fundamentally involves comparingdifferent “types” of leaders, leaving aside howa leader comes to possess the valued informa-tion or how potential followers come to know(or believe) that a person may possess valuabledata.

How does a group choose a leader? TheDewan & Myatt (2008) model and other storiesabout leader choice pose what we call the “Lifeof Brian problem,” recalling a scene in the 1979Monty Python film in which several potentialmessiahs address a skeptical crowd that is try-ing to decide which (if any) of these prophets is

18 Ahlquist · Levi


the true messiah and worth following. Althoughthis may occur, our conjecture is that the mostcommon way in which a leader emerges is notby addressing a crowd ab initio but rather byoccupying an office; the holder of this positionis expected to have better information. Thus,leadership implies organization and some regu-larization of social interaction. This conjecturealso implies that a key area for more empiricalscrutiny is those moments when organizationsand the rules and “culture” that they embodyare defined. These rules determine what knowl-edge the membership can rationally attribute tothe leader.

Systematic consideration of leadership inconcrete organizational contexts is therefore anecessity that will likely bound the availabilityof experimental opportunities needed to rigor-ously establish causal effects. As the empiricalwork from Canes-Wrone demonstrated, it isquite difficult to disentangle the causal effectof a specific individual from that of the office.Empirical strategies to determine the effect ofa leader (distinct from that of leadership) arestill in their infancy. The rich case-study litera-ture in history, political science, anthropology,management, and other disciplines is full of dataabout what leaders do and how they interactwith followers, but—as far as we can tell—a me-thodical mining of this literature is still a pipedream.

OPEN QUESTIONS ANDFUTURE DIRECTIONS

Recent contributions from political science andpolitical economy to the literature on leader-ship reveal that studying pivotal and impor-tant individuals implicates virtually all of thethorniest issues confronting social scientists to-day, from instability and chaos in multidimen-sional voting situations to the origins of pref-erences and human cognition. It is thereforehardly surprising that we have a rich but in-complete set of theoretical arguments as well asa paucity of empirical generalizations. Never-theless, in the past two decades, a more refinedand formalized discussion of what it means to

be a leader is emerging. Much of this work de-pends on the characterization of the leader as apublic and commonly observed source of infor-mation. The recent experimental work demon-strates that someone occupying this position ofleadership can affect the beliefs and behavior ofgroup members.

We have also mentioned shortcomings andchallenges in the literature, among them dif-ficulty in designing empirical studies that ac-count for the endogeneity of leaders. We con-clude by identifying two additional holes in theliterature.

The first is the role of force and coercion.Historically, social scientific studies of leader-ship have emerged out of the study of “volun-tary organizations,” e.g., firms, labor unions,churches, and social movements embedded indemocratic societies. Followers are free to with-hold contributions or leave the organization en-tirely. Little mention is made of the leaders’abilities to coerce membership participation.Indeed, Hermalin (1998, p. 1188) claims lead-ership is noncoercive: “Leadership is distinctfrom authority because following a leader is avoluntary, rather than coerced, activity of thefollowers.” Clearly, however, enforced partici-pation is part of much of political life whetherwe choose to acknowledge it or not. Some formof coercion often lurks in the background ofcurrent thinking. For example, the threat ofbeing fired from one’s job can be very power-ful motivation indeed, eliminating the need forinspirational leadership in some circumstances.Levi’s notion of “quasi-voluntary compliance”(1988, 2006) relies on the backstop of coercion.But Hermalin’s distinction is an interesting one.It seems to imply that someone occupying theposition of salience or authority—call her an“administrator” for now—has a choice over howto gain compliance from a population: eitheruse the available tools of coercion or act as a“leader” by, e.g., leading by example. Currenttheory gives us little guidance as to when wemight see an administrator relying on one orthe other.

It also seems reasonable to imagine that fol-lowers will react to coercion in different ways



depending on the larger situation at hand. Willfollowers find coercion objectionable and tryto flee the organization, or will they supportthe leader’s use of such tools? Finally, we canimagine that the availability of coercive toolsfor an administrator is ultimately endogenous.When is it in the interests of followers to pro-vide the administrator with the tools to coercethem down the road?

The second shortcoming is more normative.How can we assess the outcomes of leadershipand the processes by which leaders achieve theirobjectives? We know that leaders can manipu-late outcomes, even—or perhaps especially—indemocratic settings. However, we have yet tofully explore actual political contexts in whichmanipulation is more likely and with whatconsequences. There is, of course, importantwork on this subject. Particularly telling is theanalysis offered by Bates et al. (1998), whocombine formal theory and interpretive datato explain the fundamental shift in Serbianpolitics that helped contributed to the break-upof Yugoslavia. Slobodan Milosevic, faced with

an election he could not win, transformed thedimensions of politics. He did this by makingfear of Croat attacks on Serbs both salient andcredible. Milosevic conveyed information butalso effected a structural change. The result wasa winning coalition, which ultimately set himand his backers on the path to ethnic violenceand war. Leadership, as this case demonstrates,does not always improve aggregate welfare, andwe need to know more about the conditionsunder which it does and does not.

The use of coercion and the consequences ofmanipulation are important political phenom-ena that the literature on leadership has still toaddress. Even so, the current and growing re-search program has improved our understand-ing of coordination of group outcomes, thequality of leadership, and leaders’ capacity toaffect choices, beliefs, and policy. With furtheradvances in theory and, particularly, theoreti-cally informed empirical work, we expect evengreater illumination of what leadership means,what it does, and what we still need to learnabout it.

DISCLOSURE STATEMENT

The authors are not aware of any affiliations, memberships, funding, or financial holdings thatmight be perceived as affecting the objectivity of this review.

ACKNOWLEDGMENTS

Versions of this paper were presented at the 2010 meetings of the Southern Political ScienceAssociation, the University of Pennsylvania, and the World Bank. We thank Rick Wilson, AndyNorton, Philip Brown, Ed Campos, and Nolan McCarty for helpful suggestions. Support for theresearch for this paper was provided by National Science Foundation grants SES-0717454 andSES-0517735, the United States Study Center at the University of Sydney, and the WhiteleyCenter of the University of Washington.

LITERATURE CITED

Adorno TW. 1950. The Authoritarian Personality. New York: HarperArrow KJ. 1951. Social Choice and Individual Values. New York: WileyArrow KJ. 1974. The Limits of Organization. New York: W.W. NortonAusten-Smith D, Riker WH. 1987. Asymmetric information and the coherence of legislation. Am. Polit. Sci.

Rev. 81(3):897–918Austen-Smith D, Riker WH. 1989. Asymmetric information and the coherence of legislaton: a correction.

Am. Polit. Sci. Rev. 84(1):243–45

20 Ahlquist · Levi


Avolio BJ, Walumbwa FO, Weber TJ. 2009. Leadership: current theories, research, and future directions.Annu. Rev. Psychol. 60(1):421–49

Bala V, Goyal S. 1998. Learning from neighbors. Rev. Econ. Stud. 65:595–621Barnard CI. 1938. The Functions of the Executive. Cambridge, MA: Harvard Univ. PressBates RH, de Figueiredo R, Weingast B. 1998. The politics of interpretation: rationality, culture, and transi-

tion. Polit. Soc. 26:603–42Blanes i Vidal J, Moeller M. 2007. When should leaders share information with their subordinates. J. Econ.

Manage. Strategy 16(2):251–83Blau PM. 1955. The Dynamics of Bureaucracy. Chicago: Univ. Chicago Press. Revised ed.Blau PM. 1964. Exchange and Power in Social Life. New York: WileyBolton P, Brunnermeier MK, Veldkamp L. 2008. Leadership, coordination and mission-driven management.

NBER Work. Pap.Brehm J, Gates S. 2008. Teaching, Tasks, and Trust: Functions of the Public Executive. New York: Russell Sage

Found.Brown ME, Trevino LK. 2009. Leader–follower values congruence: Are socialized charismatic leaders better

able to achieve it? J. Appl. Psychol. 94(2):478–90Bueno de Mesquita B, Smith A, Siverson RM, Morrow JD. 2003. The Logic of Political Survival. Cambridge,

MA: MIT PressBurns JM. 1978. Leadership. New York: Harper TorchbooksCanes-Wrone B. 2006. Who Leads Whom? Presidents, Policy, and the Public. Chicago: Univ. Chicago PressCanes-Wrone B, Herron M, Schotts K. 2001. Leadership and pandering: a theory of executive policy making.

Am. J. Polit. Sci. 45(3):532–50Chattopadhyay R, Duflo E. 2004. Women as policymakers: evidence from a randomized policy experiment

in India. Econometrica 72(5):1409–43Chwe M. 1990. Why were workers whipped? Pain in a principal-agent model. Econ. J. 100:1109–21Chwe M. 1999. Structure and strategy in collective action. Am. J. Sociol. 105(1):128–56Chwe M. 2000. Communication and coordination in social networks. Rev. Econ. Stud. 67(1):1–16Chwe M. 2001. Rational Ritual. Princeton, NJ: Princeton Univ. PressCoase RH. 1937. The nature of the firm. Economica 4:386–405Coleman JS. 1963. Comment on “On the Concept of Influence.” Public Opin. Q. 27(1):63–82Coleman JS. 1980. Authority systems. Public Opin. Q. 44(2):143–63Condorcet Marquis de. 1785. Essai sur l’application de l’analyse a la probibilite des decisiones rendues a la pluralite

des voix. Paris: Imprimerie RoyaleCook KS, Hardin R, Levi M. 2005. Cooperation Without Trust? New York: Russell Sage Found.Dahl RA. 1957. The Concept of Power. New York: Bobbs-MerrillDessein W, Santos T. 2006. Adaptive organizations. J. Polit. Econ. 114(5):956–95Dewan T, Myatt DP. 2007. Leading the party: coordination, direction, and communication. Am. Polit. Sci.

Rev. 101(4):827–45Dewan T, Myatt DP. 2008. The qualities of leadership: direction, communication, and obfuscation. Am. Polit.

Sci. Rev. 102(3):351–68Dewan T, Myatt DP. 2009. On the rhetorical strategies of leaders: speaking clearly, standing back, and stepping down.

Unpublished manuscript, Dep. Gov., London School Econ. and Nuffield College, Oxford Univ.Dickson E. 2006. Rational choice epistemology and belief formation in mass politics. J. Theor. Polit. 18(4):455–

98Dickson E. 2011. Leadership, followership, and beliefs about the world: an experiment. Br. J. Polit. Sci.

In pressDirks KT, Skarlicki D. 2004. Trust in leaders: existing research and emerging issues. In Trust and Distrust in

Organizations: Dilemmas and Approaches, ed. RM Kramer, KS Cook, pp. 21–40. New York: Russell SageFound.

Eckel CC, Wilson RK. 2000. Social learning in a social hierarchy: an experimental study. Work. Pap., Dep. Polit.Sci., Rice Univ.

Eckel CC, Wilson RK. 2007. Social learing in coordination games: Does status matter? Exp. Econ. 10:317–29



Eckel CC, Wilson RK. 2009. Cooperation and status in organizations. Work. Pap., Dep. Polit. Sci., Rice Univ.Emerson RM. 1976. Social exchange theory. Annu. Rev. Sociol. 2(1):335–62Feddersen TJ, Pesendorfer W. 1998. Convicting the innocent: the inferiority of unanimous jury verdicts under

strategic voting. Am. Polit. Sci. Rev. 92:23–35Fehr E, Falk A. 1999. Wage rigidity in a competitive incomplete contract market. J. Polit. Econ. 107(1):106–34Fehr E, Kirchler E, Weichbold A, Gachter S. 1998. When social norms overpower competition: gift exchange

in experimental labor markets. J. Labor Econ. 16(2):324–51Ferreira D, Rezende M. 2007. Corporate strategy and information disclosure. Rand J. Econ. 38(1):164–84Fiorina MP, Shepsle KA. 1989. Formal theories of leadership: agents, agenda setters, and entrepreneurs. In

Leadership and Politics, ed. BD Jones, pp. 17–40. Lawrence: Univ. Press KansasFrohlich N, Oppenheimer JA, Young O. 1971. Political Leadership and Collective Goods. Princeton, NJ: PrincetonGervais S, Goldstein I. 2007. The positive effects of biased self-perceptions in firms. Rev. Finance 16:453–96Goethals GR. 2005. Presidential leadership. Annu. Rev. Psychol. 56(1):545–70Guth W, Levati MV, Sutter M, van der Heijden E. 2007. Leading by example with and without exclusion

power in voluntary contribution experiments. J. Public Econ. 91(5–6):1023–42Hargrove EC. 1989. Two conceptions of institutional leadership. In Leadership and Politics, ed. BD Jones,

pp. 57–86. Lawrence: Univ. Press KansasHeifetz RA, Linsky M, Grashow A. 2009. The Practice of Adaptive Leadership: Tools and Tactics for Changing

Your Organization and the World. Cambridge, MA: Harvard Business PressHenrich J, Gil-White FJ. 2001. The evolution of prestige: freely conferred deference as a mechanism for

enhancing the benefits of cultural transmission. Evolution Hum. Behav. 22:165–96Hermalin BE. 1998. Toward an economic theory of leadership. Am. Econ. Rev. 88(5):1188–206Hermalin BE. 2001. Economics and corporate culture. In International Handbook of Organizational Culture and

Climate, ed. S Cartwright, CL Cooper, PC Earley, pp. 217–62. Chichester, UK: WileyHermalin BE. 2007. Leadership for the long term. J. Econ. Behav. Organ. 62(1):1–19Holmstrom B. 1982. Moral hazard in teams. Bell J. Econ. 13(2):324–40Homans GC. 1958. Social behavior as exchange. Am. J. Sociol. 65(6):597–606Huck S, Rey-Biel P. 2006. Endogenous leadership in teams. J. Inst. Theor. Econ. 162:253–61Humphreys M, Masters WA, Sandbu ME. 2006. The role of leaders in democratic deliberations: results from

a field experiment in Sao Tome and Prıncipe. World Polit. 58:583–622Jackson MO. 2008. Social and Economic Networks. Princeton, NJ: Princeton Univ. PressJones BD, ed. 1989. Leadership and Politics. Lawrence: Univ. Press KansasJones BF, Olken BA. 2005. Do leaders matter? National leadership and growth since World War II.

Q. J. Econ. 120(3):835–64Keohane NO. 2010. Thinking About Leadership. Princeton, NJ: Princeton Univ. PressKobayashi H, Suehiro H. 2005. The emergence of leadership in teams. Jpn. Econ. Rev. 56:295–316Komai M, Grossman PJ. 2009. Leadership and group size: an experiment. Econ. Lett. 105(1):20–22Komai M, Stegeman M. 2010. Leadership based on asymmetric information. Rand J. Econ. 41(1):35–63Komai M, Stegeman M, Hermalin BE. 2007. Leadership and information. Am. Econ. Rev. 97(3):944–47Kreps DM. 1990. Corporate culture and economic theory. In Perspectives on Positive Political Economy, ed.

JE Alt, K Shepsle, pp. 90–143. New York: Cambridge Univ. PressLevati MV, Sutter M, van der Heijden E. 2007. Leading by example in a public goods experiment with

heterogeneity and incomplete information. J. Confl. Resolut. 51(5):793–818Levi M. 1988. Of Rule and Revenue. Berkeley: Univ. Calif. PressLevi M. 1997. Consent, Dissent and Patriotism. New York: Cambridge Univ. PressLevi M. 2005. Inducing preferences within organizations: the case of unions. In Preferences and Situations, ed.

I Katznelson, B Weingast, pp. 219–45. New York: Russell Sage Found.Levi M. 2006. Why we need a new theory of government. Perspect. Polit. 4(1):5–19Lupia A, McCubbins MD. 1998. The Democratic Dilemma: Can Citizens Learn What They Need to Know?

New York: Cambridge Univ. PressMachiavelli N. 1950 (1532). The Prince and the Discourses. New York: Modern LibraryMajumdar S, Mukand SW. 2010. The leader as catalyst: on mass movements and the mechanics of institutional

change. Mimeo, Dep. Econ., Univ. Warwick

22 Ahlquist · Levi


Marable M. 1998. Black Leadership. New York: Columbia Univ. PressMcAdam D, Tarrow S, Tilly C. 2001. Dynamics of Contention. New York: Cambridge Univ. PressMcDermott R. 2002. Experimental methods in political science. Annu. Rev. Polit. Sci. 5:31–61McKelvey RD. 1976. Intransitivities in multidimensional voting models and some implications for agenda

control. J. Econ. Theory 12:472–82McKelvey RD. 1979. General conditions for global intransitivities in formal voting models. Econometrica

47:1085–111Meidinger C, Villeval M-C. 2002. Leadership in teams: signalling or reciprocating? GATE Work. Pap.Merton RK. 1968 (1957). Social Theory and Social Structure. Glencoe, IL: Free Press. 3rd ed.Messick DM, Kramer RM, eds. 2004. The Psychology of Leadership: New Perspectives and Research. Mahway, NJ:

Lawrence ErlbaumMichels R. 1962 (1919). Political Parties. New York: Free PressMilgram S. 1974. Obedience to Authority; an Experimental View. New York: Harper & RowMiller G. 1992. Managerial Dilemmas. New York: Cambridge Univ. PressMoe T. 2005. Power and political institutions. Perspect. Polit. 3(2):215–33Moxnes E, van der Heijden E. 2003. The effect of leadership in a public bad experiment. J. Confl. Resolut.

47(6):773–95Parsons T. 1963a. On the concept of influence. Public Opin. Q. 27(1):37–62Parsons T. 1963b. On the concept of political power. Proc. Am. Philos. Soc. 107(3):232–62Plott CR, Levine ME. 1978. A model of agenda influence on committee decisions. Am. Econ. Rev.

68(March):146–60Potters J, Sefton M, Vesterlund L. 2007. Leading-by-example and signaling in voluntary contribution games:

an experimental study. Econ. Theory 33(1):169–82Przeworski A, Stokes SC, Manin B, eds. 1999. Democracy, Representation, and Accountability. New York:

Cambridge Univ. PressRiker W. 1982. Liberalism Against Populism. Prospect Heights, IL: WavelandRiker W. 1984. The heresthetics of constitution-making: the presidency in 1787, with comments on deter-

minism and rational choice. Am. Polit. Sci. Rev. 78(1):1–16Riker W. 1986. The Art of Political Manipulation. New Haven, CT: Yale Univ. PressRiker W. 1996. The Strategy of Rhetoric: Campaigning for the Ratification of the Constitution. New Haven, CT:

Yale Univ. PressRowold J, Laukamp L. 2009. Charismatic leadership and objective performance indicators. Appl. Psychol. Int.

Rev. 58(4):602–21Saiegh S. 2009. Political prowess or “Lady Luck”? Evaluating chief executives’ legislative success rates. J. Polit.

71(4):1342–56Shepsle K. 1979. Institutional arrangements and equilibrium in multidimensional voting bodies. Am. J. Polit.

Sci. 23:27–60Shepsle K, Weingast B. 1981. Structure-induced equilibrium and legislative choice. Public Choice 37:503–19Schofield N. 2006. Architects of Political Change: Constitutional Quandaries and Social Choice Theory. Cambridge,

UK: Cambridge Univ. PressSelznick P. 1957. Leadership in Administration; a Sociological Interpretation. Evanston, IL: Row, PetersonSiegel D. 2009. Social networks and collective action. Am. J. Polit. Sci. 53(1):122–38Simon HA. 1947. Administrative Behavior. New York: MacmillanStokes SC. 2001a. Mandates and Democracy: Neoliberalism by Surprise in Latin America. New York: Cambridge

Univ. PressStokes SC, ed. 2001b. Public Support for Market Reforms in New Democracies. New York: Cambridge Univ.

PressTarrow S. 1994. Power in Movement. New York: Cambridge Univ. PressThibaut JW, Kelley HH. 1959. The Social Psychology of Groups. New York: WileyTilly C. 2005. Trust and Rule. New York: Cambridge Univ. PressTucker RC. 1968. The theory of charismatic leadership. Daedalus 97(3):731–56Van den Steen E. 2004. Rational overoptimism (and other biases). Am. Econ. Rev. 94(4):1141–51



Van den Steen E. 2005. Organizational beliefs and managerial vision. J. Law Econ. Organ. 21(1):256–83Walzer M. 1973. Political action: the problem of dirty hands. Philos. Public Aff. 2(2):160–80Weber M. 1968. Economy and Society, ed. G Roth, C Wittich. New York: BedminsterWilliams B. 1981. Moral Luck. Cambridge, UK: Cambridge Univ. PressWilson RK, Rhodes CM. 1997. Leadership and credibility in N-person coordination games. J. Confl. Resolut.

41(6):767–91

24 Ahlquist · Levi

JohnAhlquist-Leadership What It Means

Documents

van der heijden

russell sage found

team production problem

dened policy space

exercise policy leadership

belief cascades

experimental setting

bd jones