Accounting for theExplanations of CEOConnpensation:Substanceand Synnbolisnn

Edward J. ZajacJames D. WestphalNorthwestern University

© 1995 by Cornell University,0001-8392/95/4002-0283/$! ,00.

Both authors contributed equally to thepaper. The helpful comments of JerryDavis, Paul Hirsch, Matt Kraatz, MarshallMeyer, Brian Uzzi, three anonymous ASOreviewers, and seminar participants atMIT and Stanford University areappreciated. We also thank RodgerVoorhies and Ann Yoo for assistance indata collection, and Linda Pike for hervaluable editorial advice.

While current debates about CEO compensation havegenerally been dominated by economic and politicalperspectives on CEO/board relations, we argue in thispaper that CEO compensation may be driven bysymbolic as well as substantive considerations. Wedevelop an interdisciplinary theoretical framework to (1)explain why alternative explanations rooted in agencyand human resource logics may be used to reduceambiguity surrounding the adoption of new incentiveplans for CEOs and (2) identify the possible structural(e.g., institutional, demographic, and economic), andinterest-based (e.g., political) factors influencing the useof such explanations. We generate and test hypothesespredicting the alternative explanations for new long-termincentive plans using data taken from proxy statementsover a 15-year period. The findings support the notionthat explanations for CEO compensation reflect bothsubstance and symbolism.*

In recent years, chief executive officer (CEO) compensationhas come under increasingly close and harsh scrutiny, bothin the academic literature and the popular press (Jensen andMurphy, 1990; New York Times, 1992), The popularbusiness press has tended to portray executivecompensation as excessive (Crystal, 1991; Time, 1993) orout of control (Business Week, 1992) and has generallysupported legislative initiatives aimed at reform (Time, 1993),Such attempts include the recent U,S, political decision toquestion the corporate tax deductibility of executive salariesabove $1 million and the efforts of some firms to limit CEOcompensation to a multiplier of entry-level-employeecompensation. Opponents of these efforts, ofteneconomists, decry the visible hand of regulation in thepresumably economic decision about the level of CEOcompensation (Jarrell, 1993) and, in the case of multiplierlimits, suggest that there is efficiency in structuringcompensation in large organizations as "tournaments" inwhich compensation at higher levels is much greater thanthat at lower levels (Lazear and Rosen, 1981),

This debate has expanded to include long-term incentivecompensation, which has become an increasingly significantcomponent of executive pay (Jarrell, 1993), Once viewed bymost observers as connoting managerial value and devotionto shareholders' interests (e,g,. Crystal, 1984), long-termincentive plans have recently been negatively portrayed asinsidious devices that serve only to enrich top managementat the expense of shareholders (Crystal, 1991; BusinessWeek, 1992), Such trenchant criticism has helped createwidespread skepticism among corporate stakeholders aboutthe motivation for long-term incentives while bolsteringlegislative efforts to curtail their use (Time, 1993), Jensenand Murphy (1990: 254-255) have suggested that firms aredeterred from making full use of long-term incentive plansfor CEOs because of the threat of "media criticism andridicule," which they view as a regrettable form of "implicitregulation," Thus debate surrounding CEO compensationreflects the tension between the economic efficiency of anideally constructed compensation contract (as agencytheorists seek to devise) versus the alleged political reality of

283/Administrative Science Quarterly, 40 (1995): 283-308

entrenched, overcompensated CEOs and weak boards. Inthis study, we argue for the need to consider a thirdperspective: the possibility that CEO compensation debatesare also driven by symbolic considerations.

From a symbolic management perspective, the concernabout whether CEOs in large U,S, corporations are overpaidor "worth every nickel they get" (Murphy, 1986: 125) is lessimportant than how firms communicate the rationalesunderlying their CEO compensation decisions. According tothis perspective, existing CEO compensation debatestypically revolve around the subjective assessment ofwhether CEO compensation decisions can be considered"justifiable" in the eyes of organizational stakeholders andappeal less to economic logic than to the subjective socialperception that some fixed amount (e,g,, $1 million) ormultiplier (e,g,, 12 times that of an entry-level person) is thelimit of what constitutes justifiable compensation for a CEO,Compensation beyond that amount could be subject todifferent interpretations by different stakeholders, somenegative and some positive. Very high CEO compensationcould be viewed either as an indication of corruption in theupper echelons of management or as a deserved reward foran extraordinary CEO, Thus, given ambiguity about thepurpose or implications of high CEO rewards, new incentivearrangements that increase the upper bound of CEOcompensation levels represent a justification problem (oropportunity) for corporate leaders. Accordingly, there may bevalue in studying not only how and how much firms decideto compensate their CEOs but also how firms explain theircompensation decisions to shareholders and other interestedconstituents, including stock analysts, governmentalregulators, governance activists, and prospectiveshareholders.

Our approach has two basic components. First, we examinethe extent to which explanations for executive incentivesreflect prevailing beliefs about the socially legitimate purposeof incentive compensation. We examine how CEOcompensation practices may be embedded in a social andorganizational context and how variation in this context(across organizations and/or across time) may result invariation in the socially legitimate explanations provided fornew CEO compensation arrangements. As Pfeffer (1981: 4)has argued, "explanations [of corporate activities are]constrained to be legitimate and acceptable in the socialcontext," Second, we examine whether explanations oflong-term incentive plans (LTIPs) reflect the interests andconcerns of corporate leaders. The LTIPs we studied addlong-term performance plans to existing arrangements (e,g,,salary and short-term bonus), and the number of firmsadopting such plans has increased significantly from themid-1970s to the late 1980s (Jarrell, 1993),

Using data taken from proxy statements over a 15-yearperiod, we seek to show that (1) when undertaking an actionthat has ambiguous meaning, such as the adoption of LTIPs,firms generally seek to provide some logic or explanation forthat action, (2) alternative explanations for that action mayexist, and (3) the construction of a theoretical frameworkthat draws on institutional, demographic, economic, and

284/ASQ, June 1995

CEO Compensation

While current conceptualizations of HRmanagement include a broader array ofconcerns, such as cost minimization andcompetency development, we focus hereon a specific dimension of this broaderliterature.

sociopolitical factors can predict when one or the otherexplanation would be used. Thus we consider how theexplanations for CEO compensation arrangements canreflect symbolic and substantive purposes.

Alternative Explanations for Long-term Incentive Plans

Given the ambiguity and controversy surrounding CEOincentive compensation, the adoption of new long-termincentive plans can present a significant justification problemfor boards and top management. There are two majoralternative explanations that may, under certaincircumstances, be viewed as appropriate rationales forincentive compensation, in particular for new LTIPs: anagency theory explanation and a human resourceexplanation. Prior research on impression managementsuggests that organizations may use media such as annualreports to address stakeholders' concerns aboutorganizational strategies or the outcomes of those strategies(Bettman and Weitz, 1983), We examine how firms use arelated media, proxy statements, to provide explanations forthe adoption of new LTIPs when announcing them, as in thefollowing extracts from recent proxy statements:

Alcoa's Board of Directors has decided to place an increasing shareof management's overall compensation at risk rather than in fixedsalaries. The new approach to compensation was recommended bythe Board's compensation committee, which is composed solely ofoutside directors. The board believes that granting stock options,performance shares and [bonuses] will create a more appropriaterelationship between compensation and the financial performanceof the company in order to increase key employees' personalfinancial identification with interests of the Company'sstockholders, (Aluminum Company of America, 1988)

The Board believes that adoption of the Plan will enhance theCompany's ability to attract and retain individuals of exceptionalmanagerial talent upon whom, in large measure, the sustainedprogress, growth and profitability of the Company depends, , , ,(AT&T, 1985)These announcements reflect two qualitatively differentexplanations for introducing LTIPs, Alcoa's explanation usesan agency logic, while AT&T's explanation uses a humanresource (HR) logic. An HR explanation highlights as animportant organizational objective the ability of anorganization to attract and retain scarce managerial talent(Milkovich and Newman, 1984; Pfeffer, 1994; Wright,McMahan, and McWilliams, 1994),'' A principal mechanismby which organizations fulfill this objective is providingcompetitive compensation arrangements that enable theorganization to attract and retain talented managers byraising the potential total compensation levels (Gomez-Mejiaand Welbourne, 1988), From an HR perspective, in whichhuman resources are critical in enhancing the firm'scompetitive position (Pfeffer, 1994), the board adopts CEOincentive plans to compete with other firms for scarceleadership talent, and an HR explanation is commonlyinvoked in explaining the administration of executive pay(e,g,, Lawler, 1990; Foulkes, 1991),

From an agency theory perspective, the primary objective ofexecutive incentive plans is to maximize shareholder wealthby aligning the interests of management (as agent) with the

2B5/ASQ, June 1995

interests of shareholders (as principal) (Jensen and Meckling,1976; Beatty and Zajac, 1994). This perspective suggeststhat, by making pay contingent on future firm performance,LTIPs mitigate the shirking behaviors (Alchian and Dennsetz,1972) or excessive perquisite consunnption (Jensen andMeckling, 1976) that enhances managennent's interests atthe expense of shareholders' welfare (Baysinger andHoskisson, 1990; Zajac, 1990; Kerr and Kren, 1992). Whilethe HR perspective ennphasizes the upside potential ofcontingent compensation, an agency perspective, with itstheoretical focus on reducing managerial shirking and thesuboptimal use of firm resources by managers who do notbear the full cost of their actions (Alchian and Demsetz,1972; Jensen and Meckling, 1976), emphasizes thedownside risk of contingent compensation.

The theoretical roots of these alternative logics can betraced historically to alternative academic discussions of therole of top managers. Early statements of the humanresource perspective on managers can be traced to Mayo's(1945) theory on organizations and their employees, whichcharacterizes organizations as arenas for natural cooperationand rejects the assumption that employees in organizationsact solely in self-interest and isolation (Perrow, 1986). Thisliterature informed later theories of compensation thatstressed the need to satisfy and thus retain valuableemployees through incentive compensation (Lawler, 1968).Most recently, HR management researchers have givenincreased attention to the implications for competitivenessof attracting and retaining those key organizationalemployees whose skills are not easily replaceable (Pfeffer,1994; Wright, McMahan, and McWilliams, 1994).

The roots of the agency perspective lie in Berle and Means'(1932) influential statement about self-interested managersin large U.S. corporations, which highlighted the problemsemerging from the growing separation of ownership andcontrol. This theme of concern for aligning managers' andowners' interests later blossomed in the academic literatureof the 1960s with the so-called managerialist school (e.g.,Marris, 1964; Williamson, 1964), which held that topmanagers pursue personal goals that are incongruent withprofit maximization for the firm. In the 1970s, what is nowcalled agency theory began to emerge from more abstractformalized analyses of principal-agent relationships and thebasic problem of delegation (e.g., Alchian and Demsetz,1972; Ross, 1973). Jensen and Meckling's (1976) statementof the agency problem and their discussion of ways tominimize agency costs in corporations is in many ways anatural extension of Berle and Means (1932).

Thus, whereas from an HR management perspective, topmanagers are valued, critical resources to be rewarded andretained (Pfeffer, 1994: 37), agency theory emphasizes theneed to minimize managerial shirking through monitoring andincentive mechanisms (Jensen and Meckling, 1976).Problems of goal congruence between the CEO and theorganization are clearly deemphasized in the formerperspective and emphasized in the latter. While thedifference between logics is relative—the HR logic iscomparatively less sensitive to issues of managerial shirking

286/ASQ, June 1995

CEO Compensation

and CEO goal incongruence and more sensitive to attractionand retention issues than the agency logic—thesecontrasting theoretical perspectives nonetheless representalternative and possibly competing explanations for CEOs'long-term incentive compensation. We suggest that anumber of specific structural and interest-based factors caninfluence which logic is used to explain LTIPs,

STRUCTURAL AND INTEREST-BASED DETERMINANTSOF ALTERNATIVE EXPLANATIONS

Prevailing Beliefs and the Content of LTIP Explanations

While the two major alternative logics to explain anorganization's decision to adopt a new LTIP for its topmanagers have both existed for many years, the legitimacyof agency and HR explanations has been changing over thelast two decades, and there has been a growingcontradiction in discussions about top management in largecorporations. On the one hand, the longstanding tendency tocelebrate individual CEOs as strong and capable corporateleaders shows no sign of waning. On the other hand, whenthe discussion of those same CEOs turns to issues of CEOcompensation, the "romance of leadership" (Meindl, Ehrlich,and Dukerich, 1985) seems to disappear, replaced by aperspective in which CEO interests are presumed to deviatefrom those of the organization (Business Week, 1992),

This discussion suggests that any organizational decision tointroduce a new CEO compensation plan is likely to beaffected by the changing institutional or social context inwhich CEO compensation decisions take place. It appearsthat the agency theoretic perspective on executivecompensation, with its emphasis on protecting owners'interests from self-interested managers, has becomeincreasingly prevalent in the business press and theorganizational literature. Recent public debate over the sizeof executive compensation packages and the accountingtreatment of long-term incentive compensation increasinglycites shareholder welfare as a primary criterion (e,g,.Business Week, 1992;T/me, 1993), Moreover, empiricalagency theory research on executive compensation hasburgeoned in recent years (see Eisenhardt, 1989; Jensenand Murphy, 1990), and organizational behavior and strategicmanagement research routinely acknowledges andfrequently incorporates agency theoretic perspectives (e,g,,Kerr and Kren, 1992; Beatty and Zajac, 1994), Even politicalperspectives on CEO compensation increasingly invokeincentive alignment (i,e,, the alignment of CEO pay andshareholder wealth) as a normative criterion or benchmark,either implicitly or explicitly, and observe deviations fromthat ideal (e,g,. Hill and Phan, 1991; Kerr and Kren, 1992),

The apparent spread of agency perspectives on executivecompensation parallels the rise of financial economic theoryin research and public policy on corporate control (Jensenand Ruback, 1983; Davis and Stout, 1992), According toDavis and Thompson (1994), this trend is closely related tothe increased activism and political influence of institutionalinvestors. In effect, large investors appear to have cooptednormative agency theory to help legitimate their political

287/ASQ, June 1995

agenda, thus contributing to and benefiting from the growthof agency theory as a donninant perspective on corporatecontrol.

Thus, while both HR and agency logics have existed fordecades, elements of the institutional context may have ledto the dominance of agency thinking about CEOcompensation over time. Davis and Thompson (1994) havereferred to this as a social movement in corporate control, inwhich social and institutional factors lead to a changingze;fge;sf with respect to corporate governance issues. Interms of LTIP explanations, the notion of protectingshareholders' interests through incentive alignment mayhave acquired institutional or symbolic value over time as arationale or explanation for the introduction of long-termincentive compensation (Meyer and Rowan, 1977; Zucker,1983), According to Meyer and Rowan (1977: 349), the useof language that provides a "legitimate account" of theorganization's activities and practices "increases thecommitment of external constituents" to the organization.Thus agency explanations may have gradually eclipsed HRexplanations for new LTIPs over time, reflecting boards'increasing emphasis on incentive alignment rather thanretaining valuable executive talent.

While the previous discussion emphasizes how the growinglegitimation of an agency perspective may have diminishedthe use of a human resource perspective, we also allow forthe possibility that both explanations may be used. Theinstitutionalization process may include a period of transitionfrom human resource to agency explanations for incentivecompensation, during which LTIP explanations have a "dualcharacter" (Zucker, 1983: 25), Just as periods of transitionbetween larger cultural systems are characterized by theintegration of old and new practices (Zucker, 1983), so mayperiods of transition in prevailing LTIP explanations reflectthe integration of declining and emerging perspectives. Thus,while the prior discussion suggests that boards are mostlikely to have used only agency explanations in recent yearsand to have used only HR explanations in early years, theymay be likely to have used both explanations during themiddle years of the study period. Taken together, then, thisdiscussion suggests the following hypotheses:

Hypothesis la (Hia): The later the date of LTIP adoption, thegreater the likelihood of agency-based explanations in LTIPannouncements.

Hypothesis Ib (Hib): The earlier the date of LTIP adoption, thegreater the likelihood of HR-based explanations in LTIPannouncements.

Hypothesis 1c (Hie): The closer the date of LTIP adoption to themiddle of the study period, the greater the likelihood of bothagency-based and HR-based explanations in LTIP announcements.

Demographic Context and the Content ofLTIP Explanations

While macro-institutional factors such as those discussedabove may provide an overarching set of social constraintson the use of alternative explanations, there may also beorganization-specific constraints. This study examines onesuch set of constraints that may be relevant in the choice of

288/ASQ, June 1995

CEO Compensation

explanations used to justify LTIPs but is rarely considered aspart of an organizational context, namely, the demographiccontext. The relevance of the demographic context can beseen in social-psychological research on performanceappraisal, which has consistently documented bias inevaluation decisions in which raters and ratees aredemographically similar (e,g,, Tsui and O'Reilly, 1989), Bothlaboratory and field research have demonstrated a positiverelationship between demographic similarity along one ormore attributes (e,g,, age, sex, race, education, or tenure)and the perceived ability, effort, and/or potential ofsubordinates (Latham, Wexley, and Pursell, 1975; Mobley,1982), More generally, abundant evidence suggests thatactual or perceived similarity enhances interpersonalattraction and liking (Homans, 1950; Byrne, 1971; Judge andFerris, 1993),

These findings are consistent with social-psychologicalresearch on intergroup relations, which has documented arobust tendency to favor in-group members in makingevaluation or allocation decisions (Messick and Mackie,1989), To the extent that demographic similarity provides asalient basis for psychological group membership (Tsui,Egan, and O'Reilly, 1992), in-group bias should produceinflated evaluations of subordinates' abilities and potentialwhen both parties are demographically similar.

Thus research in the areas of performance appraisal andintergroup relations suggests that when a large portion ofboard members are demographically similar to the CEO, theyshould be more likely to view the CEO as a valuable humanresource to be retained. As a result, increased demographicsimilarity between CEOs and board members may bepositively associated with the use of HR-managementexplanations in LTIP announcements. Under suchcircumstances, the use of an HR explanation can generatebenefits for directors on a psychological level, as well as ona social or professional level. As suggested by social identitytheory (Tajfel and Turner, 1986), directors should enjoypsychological benefits from vicarious validation of theirpersonal attributes. Affirmation of a demographically similarCEO's worth to the organization is also indirect affirmation ofthe director's own human capital, both to the director and toshareholders. Social affirmation of directors' attributes toshareholders may help justify their own compensation and,ultimately, their position on the board,

CEO-board similarity may also decrease the likelihood ofagency motivations for new LTIPs, Given that a high level ofdemographic similarity generally enhances interpersonal trust(Kanter, 1977; Useem and Karabel, 1986), the perceivedneed to direct CEO behavior and reduce shirking throughgreater incentive alignment is reduced. Moreover, given thatdemographic dissimilarity hinders communication and socialintegration (Ouchi, 1980; O'Reilly, Caldwell, and Barnett,1989), thus interfering with interpersonal monitoring of CEOdecision making, high CEO-board dissimilarity may increasethe perceived need to use LTIP adoption as an incentivealignment mechanism. Finally, the ambiguous situation of anintermediate level of CEO-board similarity—in which theCEO and board are demographically neither very similar nor

289/ASQ, June 1995

very different—may result in dual explanations. Thus, just asa firm's institutional context can influence the content ofLTIP logics, so too can a firm's demographic context, byinfluencing the beliefs of corporate actors about the purposeof new LTIPs,

The basis for perceived similarity or psychological groupmembership can be a characteristic itself or the underlyingskills, philosophy, or behavioral style that the characteristicindicates. Thus, for instance, directors may recognize CEOsas fellow finance-types, or alternatively, perceived similaritymay derive from recognition of the CEO's commonphilosophies about strategic goals and objectives, as indexedby similarity in functional backgrounds (Hambrick and Mason,1984; Bantel and Jackson, 1989; Tsui, Egan, and O'Reilly,1992), Similarity in educational background may also indicatecommon socioeconomic status, as well as similar leadershipand communication styles (Collins, 1979; Hambrick andMason, 1984; Useem and Karabel, 1986), Considerabletheory and evidence also suggest that age provides a salientbasis for group identification (e,g,, Stangor et al,, 1992) andthat variation in age predicts a variety of attitudinal andbehavioral differences, including risk orientation andapproaches to strategic decision making (Hitt and Tyler,1991), Finally, there is growing evidence that peopleperceive similarity on the basis of multiple social features, sothat in-group bias is more likely when people share multiplegroup memberships (Stangor et al,, 1992); accordingly, wealso examine the independent effect of similarity acrossmultiple demographic attributes. This discussion suggeststhe following specific hypotheses:

Hypothesis 2a (H2a): The less demographic similarity there isbetween the CEO and the board (measured in terms of functionalbackground, education, and age), the greater the likelihood ofagency-based explanations in LTIP announcements.

Hypothesis 2b (H2b): The greater the demographic similaritybetween the CEO and the board, the greater the likelihood ofHR-based explanations in LTIP announcements.

Hypothesis 2c (H2c): When the CEO and the board aredemographically neither very similar nor very different, bothagency-based and HR-based explanations are likely to be used inLTIP announcements.

Performance Context and the Content ofLTIP Explanations

We also consider another set of organization-specificconstraints, namely, how an organization's periformancecontext may influence the selection of an agency or HRexplanation when introducing new LTIPs, Organizations needto select explanations that are plausible (Morrill, 1991) inlight of current firm performance. Thus, for example, byusing an agency explanation when performance is poor,boards implicitly acknowledge a lack of convergencebetween stockholders' and CEOs' interests, while alsoimplying that the board has rectified this problem throughLTIP adoption. This logic emphasizes that CEOs will now bein the same situation as shareholders: If firm performancedoes not improve, and shareholders continue to suffer as aresult, the CEO will also suffer,

290/ASQ, June 1995

CEO Compensation

Under conditions of relatively high firnn performance, bycontrast, boards nnay use HR-based explanations, givenapparent evidence of the value of their executive's talentand the innportance of retaining it. In effect, good firmperformance provides an opportunity to invoke the romanceof leadership (Meindl, Ehrlich, and Dukerich, 1985) as acredible justification for long-term incentive compensation,implicitly associating LTIPs with an optimistic perspective oncorporate leadership. When performance is neither high norlow, both explanations are credible, given that performanceis ambiguous. Thus, the firm's performance context mayrepresent an important constraint on the content of LTIPjustifications. We thus hypothesize:

Hypothesis 3a (H3a): The poorer a firm's performance, the greaterthe likelihood of agency-based explanations in LTIPannouncements.

Hypothesis 3b (H3b): The higher a firm's performance, the greaterthe likelihood of HR-based explanations in LTIP announcements.

Hypothesis 3c (H3c): The closer a firm's performance is to anaverage level, the greater the likelihood of both agency-based andHR-based explanations in LTIP announcements,

CEO vs. Board Power and the Content ofLTIP Explanations

From an interest-based perspective, influential organizationalleaders may seek to manipulate actively externalconstituents' perceptions of CEO compensation. This isreflected in how the relative power of the CEO vis-a-vis theboard of directors may affect the use of alternativeexplanations for the introduction of LTIPs, From a powerperspective, explanations for LTIPs may be contested,reflecting and perhaps reinforcing the relative power ofcorporate actors (Friedland and Alford, 1991), While CEOsmay favor explanations that reflect romanticized conceptionsof corporate leaders, thus accentuating their perceived valueto the firm, powerful boards may favor accounts thatproclaim their control over management. In effect, HRexplanations of LTIPs afford reputational benefits to CEOs,while agency explanations provide reputational benefits toboard members. We examine the impact of four potentialsources of board power on the content of LTIP justifications:separation of the CEO and board chair positions, the portionof the board composed of outsiders appointed before theCEO, low CEO tenure, and the relative stock ownership ofthe board and CEO,

Separation of the CEO and board chair positions.Managerial power theorists, agency theorists, and advocatesof board reform have all argued that creating separate CEOand board chair positions enhances the board's independentmonitoring capacity and curtails managerial entrenchment(Rechnerand Dalton, 1991; Finkelstein and D'Aveni, 1994;Beatty and Zajac, 1994), In contrast, CEOs holding bothpositions simultaneously are thought to possess greaterformal authority over board members and heightenedinformal stature (Patton and Baker, 1987; Harrison, Torres,and Kukalis, 1988), Thus when these positions areseparated, boards are better able to frame explanations fornew LTIPs in terms of controlling CEO behavior. Conversely,

291/ASa June 1995

when CEOs are also board chairs, they are better able tosignal that they, as uniquely talented leaders, deserveminimal board interference in managing corporate affairs(Finkelstein and D'Aveni, 1994); thus they are more likely toframe new LTIPs as a means of retaining scarce leadershiptalent.

Appointment of outsiders. Several authors have arguedthat, through control of the director-nominating process,CEOs are able to select outsiders with whom they have apersonal relationship or who are otherwise sympathetic tothem (Mace, 1971; Fredrickson, Hambrick, and Baumrin,1988; Wade, O'Reilly, and Chandratat, 1990), Like insiders,these directors may feel beholden to CEOs for their position.Thus when CEOs have appointed a relatively large portion ofoutside directors to the board, HR explanations are likely tobe used for new LTIPs; in contrast, when a relatively largeportion of the board is composed of outsiders appointedbefore the CEO was appointed, agency explanations aremore likely to be used in LTIP announcements,

CEO tenure. Several studies have hypothesized that boardcontrol over management diminishes as CEO tenureincreases (Finkelstein and Hambrick, 1989; Singh andHarianto, 1989), As CEO tenure increases, CEOs develop a"personal mystique or patriarchy" (Finkelstein and Hambrick,1989: 124), resulting in sanctions against questioning theCEO's authority, CEOs are also thought to coopt boards overtime by appointing sympathetic outsiders (Wade, O'Reilly,and Chandratat, 1990), Thus when CEO tenure is low,boards should have greater influence over the symboliccontent of LTIP explanations, and agency-based explanationsare likely, while high CEO tenure should increase thelikelihood of HR explanations.

Relative director and CEO stock ownership. From anagency perspective, board members' stock ownership moreclosely aligns the interests of board members with theinterests of shareholders, creating an incentive for the boardto exercise greater control over the CEO compensation-setting process. Moreover, voting rights afford additionalpower to owner-directors, and this power increases with theportion of total shares held (Zaid, 1969), Thus higher stockownership may give directors greater relative power vis-a-visthe CEO and therefore lead to greater use of agencyexplanations and less use of HR explanations whenannouncing new LTIPs, Because stock ownership is also animportant source of power for the CEO, we examine theeffect of relative stock ownership (i,e,, director stockownership divided by CEO ownership) on the use of eachlogic (Finkelstein, 1992),

For each of these measures of power, we also consider thesituation in which power is relatively balanced between theCEO and the board. Under such circumstances, both partiesmay be sufficiently powerful to influence the content of LTIPexplanations. The presence of both explanations in LTIPannouncements thus may represent a compromise or trucebetween CEOs and boards. This discussion suggests thefollowing hypotheses:

292/ASQ, June 1995

CEO Compensation

Hypothesis 4a (H4a): The greater the power of the board(measured in terms of separate CEO and chair positions, theproportion of the board composed of outsiders appointed beforethe CEO, a CEO with relatively low tenure, and high director stockownership), the greater the likelihood of agency-based explanationsin LTIP announcements.

Hypothesis 4b (H4b): The greater the power of the CEO, thegreater the likelihood of HR-based explanations in LTIPannouncements.

Hypothesis 4c (H4c): The more balanced the distribution of powerbetween the board and the CEO, the greater the likelihood thatboth agency-based and HR-based explanations will be used in LTIPannouncements,

METHOD

Data

The initial sample for this study consisted of all LTIPadoptions between 1976 and 1990 among firms listed in the1976 Forbes 500 or Fortune 500 indexes. The Forbes 500uses multiple lists whose overlap depends on the specificsize measure used. This study used those firms thatappeared on at least two lists. We chose the year 1976 toapproximate the onset of LTIP adoption, defined here as theaddition of a new performance plan that is aimed atproviding multiyear performance incentives, such asperformance shares or performance units, to a CEO'scompensation contract (Westphal and Zajac, 1994), Weexcluded amendments or updates to existing plans andadoptions for which sufficient proxy statements wereunavailable to identify the specific year of adoption. Weincluded cases only if complete demographic data wereavailable for at least three-quarters of the outside directors ineach year. This procedure yielded a final sample of 352firms. Two-sample r-tests revealed no significant differencesin size (measured as sales and number of employees) orperformance (measured as return on assets and total stockreturns) between the initial and final samples.

We collected data on performance and CEO influence(tenure, separate CEO and board chair, and the appointmentof outsiders) for the years 1976 to 1990, inclusive. Twoindicators of CEO influence (CEO tenure and separate CEOand board chair) were collected for each year, while data onoutsiders were gathered at three-year intervals beginning in1976, Stock ownership data were collected for the year ofadoption. Information to measure CEO influence wasobtained from both proxies and Standard and Poor's Registerof Corporations, Directors, and Executives. Demographicinformation was obtained from the Dun & BradstreetReference Book of Corporate Managements and from IMJO'SWho in Finance and Industry. Data from Standard and Poor'sCOMPUSTAT service and the Center for Research inSecurity Prices (CRSP) were used to calculate performanceand size measures.

Dependent Variables

To determine the presence or absence of variousexplanations for new LTIPs, we conducted a contentanalysis of proxy statements in the year of adoption (Holsti,1968; Weber, 1985), Boards commonly announce the

293/ASa June 1995

adoption of new LTIPs in a separate section of the proxystatement (Westphal and Zajac, 1994). This announcennenttypically includes a detailed description of the plan's features(e.g.. administration, eligibility, award types), which may bepreceded by an introductory section outlining the reasons foradopting the plan. When explanations are provided, they aretypically included in this introductory portion of theannouncement. Nonetheless, we carefully read each proxystatement from the beginning of the compensation sectionforward.

Since we anticipated little ambiguity in the coding process,we decided not to develop an exhaustive rule bookspecifying the appropriate categorization of every possiblephrase or combination of phrases, a strategy that canjeopardize the content validity of the coding scheme (Holsti,1968). Instead, we simply provided coders with summarydescriptions of agency and HR perspectives, together with ashorter list of key concepts characterizing each theory, andspecific coding instructions. The summary descriptions werevery similar to the paragraphs above describing eachperspective shown earlier. While some of these conceptsmay appear abstract and theoretical, boards typically usedterminology from the theories in explaining the plan (e.g.,"align" pay with shareholder returns, or "attract and retain"executive talent).

Three people independently coded the relevant sections ofall proxy statements that included some explanation for thenew plans. Two of the coders were doctoral students inbusiness, and the third was a demographically differentundergraduate. The coding itself consisted of two sequentialand dichotomous choices (Holsti, 1968): (1) Was an agencytheory explanation used? (2) Was a human resourcemanagement explanation used? Coders were instructed toanswer the first question before proceeding to the second.Based on these classifications, we created threedichotomous dependent variables (agency explanation, HRexplanation, and agency and i-iR explanations), coded 1 if therelevant explanation(s) were used in the proxy statement,and 0 otherwise. Thus the unit of analysis is the entire proxystatement. Coders also noted whether some otherexplanation was used. Analysis of prenegotiation intercoderreliability yielded Pearson correlation coefficients rangingfrom .903 to .972, and the rate of intercoder agreement was95 percent for both explanations, suggesting minimalambiguity in the coding scheme.

Independent Variables

Board/CEO influence. Separate CEO and board chair is abinary variable, coded 1 if a CEO is not also chairman of theboard, and 0 otherwise. Appointment of outsiders indicatesthe portion of the board composed of outside directorsappointed after the CEO. CEO tenure was measured inyears. Finally, relative directors' stock ownership wasmeasured as the percentage of common stock owned bynon-CEO directors divided by the percentage held by alldirectors, including the CEO.

Firm performance. The impact of performance on thecontent of LTIP explanations was measured by two

294/ASQ, June 1995

CEO Compensation

variables. Total stock returns was used as a nnarket-basedmeasure of firnn performance, calculated as capital gainsplus dividends accrued/paid during the year, divided by shareprice at the beginning of the year. Return on assets wasused as an operating measure of performance. Wemeasured performance in the year of adoption. Results werenot sensitive to altemative operationalizations ofperformance, such as average stock returns or return onassets over the past three or five years.

Demographic similarity. In measuring functionalbackground similarity, educational degree similarity, andeducational affiliation similarity between the CEO and theboard, we applied a variant of Blau's (1977) index ofheterogeneity, defined as (P,)̂ , where P, is the proportion ofCEO-board-member dyads sharing the /th category. Thus,for functional background, this measure indicates thesquared proportion of CEO-board-member dyads in whichboth individuals have primary experience in the samefunctional area. Following Hambrick and Mason (1984) andMiles and Snow (1978), we consolidated the variousfunctional backgrounds into three core areas: outputfunctions, which include marketing and sales; throughputfunctions, which include operations, research anddevelopment, and engineering; and peripheral (or staff)functions, comprising law, finance, and accounting. In aseparate model, we used a more fine-grained classificationscheme consisting of nine different categories (Bantel andJackson, 1989) and found that the results were substantivelyunchanged. We opted to use Hambrick and Mason's (1984)classification scheme in the final models, however, givenempirical evidence that this particular measure of functionalbackground explains variation in generic strategy (Chagantiand Sambharya, 1987).

For educational degree, this measure represents the portionof such dyads in which both individuals have the same typeof degree. Degree type was measured as the highestobtained university degree and was divided into fivecategories; arts, sciences, engineering, management oreconomics, and law (Wiersema and Bantel, 1992). Whenspecialities were not listed, B.S, and M,S, degrees wereclassified as science degrees. Educational affiliation similaritysimply indicates the portion of CEO-board-member dyads inwhich both individuals attended the same educationalinstitution, at the undergraduate or graduate level.

Age similarity was measured with an analog of the euclideandistance measure (i,e,, the coefficient of variation) commonlyused in research on organizational demography (O'Reilly,Caldwell, and Barnett, 1989);

where S, is the CEO's age, Sj indicates the age of directory(excluding the CEO), and n represents the number ofdirectors. This measure was converted to an indicator ofsimilarity by subtracting each firm's coefficient from thehighest value in the sample. Finally, we measured similarity

295/ASQ, June 1995

on multiple dimensions as the portion of CEO-board-member dyads sharing the same demographic category forat least three of the four dimensions. To create this variable,we first created a dichotonrious measure of age similarity,coded as 1 for a given dyad if the individuals differed in ageby less than one standard deviation, and 0 otherwise.

Other independent variables. To test HI , we created avariable indicating time of adoption, with values ranging from1 (adoption in 1976) to 15 (adoption in 1990), In addition,given that larger firms may be relatively visible to externalstakeholders (Jensen and Murphy, 1990), and externalvisibility may be differentially related to the use of agencyand HR explanations, log of sales was included as a controlvariable. Finally, to test hypotheses relating intermediatelevels of the variables measuring time, firm performance,relative CEO-board power, and demographic similarity to theuse of both explanations, we converted each independentvariable into raw deviation form (i,e,, \X - X\). While ourtheoretical predictions refer to all large U,S, corporationsadopting LTIPs, to ensure that our findings are not materiallyaffected by industry-specific differences, we also controlledfor industn/ by including dummy variables for all two-digitStandard Industry Classification codes represented in thesample and found that the results were substantivelyunchanged.

Data Analysis

We used logistic regression models to analyze the likelihoodof firms' using agency, HR management, or bothexplanations. Ordinary least squares (OLS) analysis isinappropriate when the dependent variable is binary,because OLS assumes a linear additive model with normallydistributed error terms, while the true probability model isnonlinear with binomially distributed errors (Aldrich andNelson, 1984), The logistic function we used takes thefollowing form:

log[P,/(i - P,)] = b^,,,

where P, is the probability of observing the relevantexplanation, X,,.s are independent variables, and b̂ -s are theestimated coefficients, P, is defined as:

such that P, increases monotonically with b/^n, and canassume any value between zero and one.

We analyzed the likelihood of using agency explanations, HRexplanations, or both explanations in three separate logitmodels, corresponding with the a-b-c ordering of thehypotheses. Thus, for instance, the agency explanationmodel assesses the likelihood that an agency explanationwill be used rather than a different explanation or noexplanation at all. The coefficients from this model indicatethe increase in log odds (i,e,, the logit) that the firm will usean agency explanation for each one-unit increase in theindependent variable (Allison, 1984), Because the use of twoexplanations (i,e,, agency and HR) was coded 1 in all threeanalyses, the descriptive statistics showing the percentageof agency, HR, and dual explanations exceed 100 percent,

296/ASQ, June 1995

CEO Compensation

This modeling approach represents a very conservative testof the hypotheses predicting the use of only agency or onlyHR explanations. In separate unreported logistic analyses,we also modeled the probability of using only an agencyexplanation and the probability of using only an HRexplanation by coding dual explanations as 0 in these twoanalyses. Predictably, the results were stronger than thosereported here in Tables 2 and 3, below. Also, to increase ourconfidence in the results, we conducted a multinomiallogistic regression analysis (Aldrich and Nelson, 1984;Greene, 1993) in which LTIP explanations were divided intothree categories: (1) agency only, (2) HR only, and (3) agencyand HR, This model estimates separate coefficients for eachexplanation type, given a base category of no justification.The findings were consistent with the results reported here,

RESULTS

Table 1 provides the means, standard deviations, andbivariate correlations for all variables in the study. As Table 1shows, nearly half (47 percent) of all firms in the sampleused an agency explanation in announcing new LTIPs, and aslight majority (51 percent) used HR-managementexplanations. Moreover, 19 percent of these firms used bothexplanations.

The results of logistic regression analyses predicting thelikelihood of agency, HR-management, or dual explanationsare provided in Tables 2, 3, and 4, respectively, Hiapredicted that later adopters would be more likely to useagency explanations. The findings in Table 2 support thishypothesis: Time of adoption is positively and significantlyrelated to the likelihood of agency explanations. The resultsprovided in Table 3 also support Hib: The later the year ofadoption, the lower the likelihood of HR explanations. Finally,the results shown in Table 4 strongly support Hie, whichpredicted that adoption nearer the middle of the study periodis likely to lead to the use of dual explanations. The closerthe time of adoption to the mean value, the greater thelikelihood of the use of both agency and HR explanations.Because the independent variables were converted to rawdeviation scores, statistically significant, negative coefficientsprovide support for the hypotheses,

H2a and H2b related measures of demographic similaritybetween the CEO and the board to the use of agency andHR explanations in LTIP announcements. The results inTables 2 and 3 generally support these hypotheses. Inparticular, four of the five measures of demographicsimilarity (similarity in functional background, educationaldegree, educational affiliation, and across multipledimensions) are positively and significantly related to the useof HR explanations and negatively related to the use ofagency explanations in LTIP announcements, H2c is notsupported by the results in Table 4: The use of dualexplanations is not more likely at intermediate levels ofCEO-board demographic similarity for four of the fivesimilarity measures.

The findings in Tables 2 and 3 strongly support H3a andH3b, respectively. Both measures of firm performance are

297/ASQ, June 1995

Table 1

Descriptive Statistics and Pearson Correiation Coefficients {N =

Variable

1. Time of adoption2. Total stock returns3. Return on assets4. Separate CEO and board chair5. CEO tenure6. Appointnnent of outsiders7. Reiative director and CEO ownership8. Functional background similarity9. Educational degree similarity

10. Educational affiliation similarity11. Age similarity12. Similarity on multiple dimensions13. Agency explanation14. Human resource explanation15. Agency and HR explanations16. Log of sales

Mean

9.017.345.03

.167.99

.29

.41

.52

.38

.1219.35

.11

.47

.51

.196.86

S.D.

4.3916.555.78

.375.01

.22

.34

.31

.26

.143.19

.10

.41

.40

.39

.98

352)

1

.04

.05

.08- .17- .05

.07

.19

.33

.14

.20

.12

.26- .22

.04

.05

2

.28

.07- .06- .08

.17- .10

.03

.02- .11- .08- .24

.33

.02

.09

3

.21- .08- .04

.04- .11

.02- .06- .16- .10- .18

.26- .06

.03

4

- .20- .23

.18

.15

.24

.08

.07

.03

.32- .15

.01

.05

5

.41- .18- .05- .08- .05- .02- .04- .28

.16- .06

.11

negatively related to the likelihood of agency explanationsbut positively related to the likelihood of HR-nnanagennentexplanations in LTIP announcements. High-performing firmstend to use HR explanations, while poorly performing firmsrely on agency explanations to justify new LTIPs. Moreover,as shown in Table 4, deviations from average performancelevels are negatively associated with the use of dualexplanations. Thus, consistent with H3c, firms are morelikely to use both explanations at intermediate levels of firmperformance.

H4a and H4b predicted that greater board power increasesthe likelihood of agency explanations and decreases thelikelihood of HR explanations. In general, these hypothesesare strongly supported in Tables 2 and 3. Consistent withH4a, all four measures of board power over the CEO(separate CEO and board chair, relative director and CEOstock ownership, low CEO tenure, and few outside directorsappointed by the CEO) are positively associated with the useof agency explanations. A similar pattern of findingsemerges in Table 3, supporting H4b: Three of the fourmeasures of CEO power over the board (combined CEO andboard chair positions, low relative director and CEO stockownership, and outsiders appointed after the CEO) arepositively associated with the use of HR explanations,suggesting that relatively high CEO power over the boardincreases the use of HR explanations in LTIPannouncements. The results in Table 4 also afford strongsupport for H4c: The use of dual explanations is more likelyat intermediate levels of board power, as indicted by theappointment of outside directors (significant at p ^.10), CEOtenure, and relative director and CEO stock ownership(separate CEO and board chair, a categorical variable, wasexcluded from this specific analysis). The implications ofthese results are discussed below.

We also examined whether firms may be using agency andHR logics for LTIPs in a more straightforward, nonsymbolicway to report differences in how the plans are actually used.HR explanations may reflect the use of LTIPs to increase

298/ASQ, June 1995

CEO Compensation

Table 1 {cor)tir)uedl

10 11 12 13 14 15

14110910140718130509

,22,17,03,06,08,23

-.24,03,04

,42,45,25,46

-,19,30,08,17

,52,31,53

-,17,22,05,18

,36,41

-,15,18

-,01,03

,55-,30,20,04

-,10

-,17,16,04,10

-,11,41,03

,38,02 -,01

pay level, while agency explanations may reflect the use ofLTiPs to increase pay contingency. Similarly, high-performingfirms may simply use LTIPs as a way to provide additionalcompensation, while low-performing firms may use them toincrease compensation contingency, measured as the valueof long-term incentives granted to the CEO in a particularyear divided by total compensation (i,e,, salary plus annualbonus plus the value of long-term incentives). We found noevidence supporting these alternative assumptions.Regression analyses of change in compensation level andcontingency following adoption revealed that firms usingagency explanations did not increase compensationcontingency after LTIP adoption any more than those firmsnot using agency explanations and that firms using HRexplanations did not increase the total compensation levelafter LTIP adoption any more than those firms not usingsuch explanations. In addition, we found that low-performing

Table 2

Logistic Regression Anaiysis Predicting the Use of AgencyExpianations (A/ = 352)*

UnstandardizedIndependent variable coefficients

1, Time of adoption ,096 (,047)*2, Total stock returns -,021 (,010)*3, Return on assets -9,997 (3,369)*~4, Separate CEO and board chair 1,150 (,446)~5, CEO tenure - ,090 (,049)*6, Appointment of outsiders -2,663(1,218)*7, Relative director and CEO ownership 2,891 (1,148)'*8, Functional background similarity -1,637 (,536)*^9, Educational degree similarity -1,784 (,678)**

10, Age similarity - ,059 (,050)11, Educational affiliation similarity -2,619(1,364)*12, Similarity on multiple dimensions -11,124(4,013)**13, Log of sales -,183 (,168)

Constant 2,773(3,416)

Chi-square 136,78***

* p < ,05; **p s ,01; ***p s ,001; f-tests are one-tailed for hypothesizedeffects, two-tailed for control variables,

" Standard errors are in parentheses,

299.'ASQ, June 1995

Table 3

Logistic Regression Anaiysis Predicting the Use of Human ResourceExpianations (/V = 352)*

UnstandardizedIndependent variable coefficients

1, Time of adoption - ,092 (,043)*2, Total stock returns ,017 (,008)*3, Return on assets 5,979(3,152)*4, Separate CEO and board chair - ,981 (,385)**5, CEO tenure ,049 (,040)6, Appointment of outsiders 5,549(1,565)***7, Relative director and CEO ownership -2,813(1,286)*8, Functional background similarity 1,023 (,489)*9, Educational degree similarity 1,236 (,601)*

10, Age similarity - ,044 (,045)11, Educational affiliation similarity 3,582 (1,374)**12, Similarity on multiple dimensions 7,393(3,227)*13, Log of sales ,372 (,152)*

Constant -2,079(2,147)

Chi-square 113,58***

* p s ,05; **p s ,01; ***p £ ,001; t-tests are one-tailed for hypothesizedeffects, two-tailed for control variables,

• Standard errors are in parentheses.

Table 4

Logistic Regression Analysis Predicting the Use of Both Agency andHuman Resource Expianations [N = 352)*

UnstandardizedIndependent variable coefficients

1, Time of adoption -,247 (,091)**2, Total stock returns -,008 (,004)*3, Return on assets -10,002 (4,464)*4, CEO tenure - ,211 (,095)*5, Appointment of outsiders - 1,681 (1,262)6, Relative director and CEO ownership -3,062 (1,540)*7, Functional background similarity 1,296(1,265)8, Educational degree similarity -1,101(1,395)9, Age similarity ,092 (,088)

10, Educational affiliation similarity -3,868(1,690)*11, Similarity on multiple dimensions -2,516(2,014)12, Log of sales -,175 (,194)

Constant 5,809(1,836)***

Chi-square 76,84***

* p s ,05; **p s ,01; ***p s ,001; t-tests are one-tailed for hypothesizedeffects, two-tailed for control variables,

* Standard errors are in parentheses. Each of the independent variables exceptlog of sales was transformed to equal the absolute value of the raw deviation.

firms did not increase compensation contingency after LTIPadoption any more than high-performing firms and thathigh-performing firms do not increase the totalcompensation level after LTIP adoption any more thanlow-performing firms. These findings suggest that LTIPs arenot used differently by firms providing different explanationsand that the choice of explanations provided may be drivenmore by symbolic considerations.

DISCUSSION

The present study is distinctive in at least three ways. First,it highlights how a symbolic management perspective

300/ASQ, June 1995

CEO Compensation

provides a relevant new perspective in discussing CEOcompensation contracts. Second, it explains why alternativeexplanations may exist to justify the adoption of newincentive plans for CEOs. Third, it uses an interdisciplinaryapproach that shows how multiple dimensions of anorganization's macro and micro context, along withmanagerial self-interest, can predict the use of particularexplanations. The findings suggest that (1) organizations usealternative explanations to justify the adoption of CEOs'long-term incentive plans, and (2) these alternativeexplanations are not chosen randomly but can be predictedby variables capturing the structural (e.g., institutional,demographic, and economic) and interest-based (e.g.,political) factors facing these organizations.

The descriptive findings confirm that boards frequentlyexplain new LTIPs in terms of rewarding and retainingscarce leadership talent, and/or controlling executivebehavior by aligning executive pay more closely withshareholders' interests. Thus, in explaining the purpose ofnew LTIPs to shareholders, boards appear to drawconsciously or unconsciously on HR-management and/oragency theory logics as potentially legitimate institutionallogics (Friedland and Alford, 1991) for incentivecompensation. These findings also indicate that there is nouniform boilerplate language used in introducing LTIPs;rather, different logics are used, with similar phrasing withinlogics. Although one could argue that if consultants or otherthird parties sometimes write these documents, directorsand top managers have little influence over their content,even when corporate leaders exercise little direct influenceover the writing process itself they can exercise indirectinfluence (Mizruchi, 1983) by dictating the climate withinwhich communications are formulated (Arendt, 1986) or byselecting third parties with similar orientations towardincentive compensation.

The multivariate results address the institutional,demographic, economic, and sociopolitical factors leadingorganizations to use one explanation and/or the other injustifying new LTIPs. For example, we found that agencytheoretic perspectives have become more prevalent in LTIPexplanations over time, while HR-management perspectiveshave become less common. This result suggests thatagency theoretic concepts such as incentive alignment andcontrol may have acquired institutional or symbolic valueover time as explanations for incentive compensation (Meyerand Rowan, 1977; Zucker, 1983). The analyses also showthat firms are most likely to use both agency and HRexplanations near the middle of the study period, suggestingthat institutional logics have a "dual character" duringperiods of transition between dominant logics (Zucker, 1983:25).

The second set of findings shows that demographicsimilarity between the CEO and the board increases thelikelihood of HR explanations while decreasing the likelihoodof agency explanations in LTIP announcements. It appearsthat board members perceive demographically similar CEOsas valuable human resources to be rewarded and retainedand that LTIPs are explained in this way. In contrast, board

301/ASQ, June 1995

nnennbers apparently view out-group CEOs as agencyproblenns rather than valued human resources. Drawing fromsocial psychological research on intergroup perceptions,Kramer (1991: 210) argued that "ingroup members tend toperceive outgroup members as less trustworthy, lesshonest, and less cooperative than members of their owngroup," Thus CEOs with different functional specializationsor educational backgrounds may be perceived as lesstrustworthy than CEOs with similar philosophies, skills, orsocial backgrounds (Kanter, 1977; Useem and Karabel,1986), In addition, the findings indicate that demographicsimilarity across multiple dimensions has an independenteffect on the use of different explanations, after controllingfor the effect of similarity on individual dimensions. Thisfinding is consistent with recent evidence that individualsuse multiple social features to assess similarity, increasingthe likelihood of in-group bias when individuals sharemultiple group memberships (Stangor et al,, 1992), Moregenerally, the findings suggest that the demographic contextof organizations—by introducing cognitive biases—can exerta constraining effect on the content of impressionmanagement directed toward shareholders.

The third set of findings suggests that the content of LTIPexplanations is also influenced or constrained by theperformance context in which LTIPs are introduced. Whenperformance is poor, the observed tendency of firms to usean agency explanation implicitly suggests both the problemand solution: lack of incentive alignment to be remedied bythe adoption of a new LTIP, This explanation alsoemphasizes that CEOs will suffer along with shareholders ifperformance does not improve. Similarly, DeAngelo andDeAngelo (1991: 6) found, controlling for firm performance,that companies were less likely to pay bonuses to topmanagement during union negotiation years. They concludedthat "top management's financial sacrifices appear to besymbolic actions which indicate that they too will bearpersonally significant costs during the firm's financialdifficulties," When performance is high, boards implicitlyattribute this success to the CEO by affirming the value ofexecutive talent and the importance of retaining it (Staw,McKechnie, and Puffer, 1983; Salancik and Meindl, 1984),Finally, the analyses suggest that both explanations areplausible when performance is average or ambiguous.

The fourth and final set of findings deals with interest-baseddeterminants of LTIP explanations, namely, the relativepower of CEOs and boards. The results show that boardinfluence over the CEO increases the likelihood of agencyexplanations, while decreasing the likelihood of HRexplanations. Thus, consistent with a political perspective onimpression management (Pettigrew, 1985), it appears thatthe content of LTIP explanations may be "contested" amongorganizational actors, reflecting and perhaps reinforcing theexisting distribution of power within the firm (Friedland andAlford, 1991), While powerful CEOs advertise their value tothe firm and rationalize their power over the board byfavoring LTIP explanations that reflect romanticizedconceptions of corporate leaders, powerful boards favoraccounts that emphasize board control over management,

302/ASQ, June 1995

CEO Compensation

Consistent with this perspective, the results also suggestthat, when board control over management is neitherparticularly high nor low, firms are most likely to use bothagency and HR explanations. It appears that under suchcircumstances both parties are sufficiently powerful toinfluence the content of LTIP explanations. In effect, thepresence of both explanations in LTIP announcements mayrepresent a compromise between management's andshareholders' interests.

This study extends prior research on the role of substanceand symbolism in executive compensation (Westphal andZajac, 1994), First, we examined how CEO compensationpractices may be significantly influenced by the social andorganizational context and how variation in this contextacross organizations and/or across time resulted in variationin the explanations provided for new CEO compensationarrangements. This aspect of our approach is consistent withPfeffer (1981: 4), who noted that explanations of corporateactivities "are constrained to be legitimate and acceptable inthe social context," By choosing explanations within theseconstraints, organizations, as Meyer and Rowan (1977: 50)suggested, "provide an account of activities that protectsthe organization from having its conduct questioned," In thepresent study, institutional logics or accounts were providedto enhance the legitimacy of new formal structures oractivities (i,e,, LTIPs), The study also extends this literatureby specifying how multiple sources of constraint, bothinstitutional and organization-specific factors, may affectwhich explanations are provided.

Second, the study's findings also support the moreinterest-based perspective on organizational communication,wherein poweriul corporate leaders seek to encouragefavorable interpretations of organizational actions and thusenhance their professional reputations (Elsbach and Sutton,1992), It appears that organizational participants exercisepower not only "on the level of substantive action" (e,g,, todetermine the form of executive compensation contracts),but also "on the expressive or symbolic level" to ensure thecommitment and support of organizational stakeholders(Pfeffer, 1981: 1), Future research might extend the presentstudy by examining whether political conflict and debateover compensation issues relevant to employees at lowerlevels of the organization, such as the minimum wagedebate, are also carried out on both symbolic andsubstantive levels.

Moreover, this research makes several contributions to thegrowing literature on organizational impression management.First, while most research in this area has analyzed pressstatements or letters to shareholders (e,g,, Staw,McKechnie, and Puffer, 1983; Salancik and Meindl, 1984;Marcus and Goodman, 1991; Elsbach, 1994), the presentstudy describes a unique vehicle for symbolic management:LTIP explanations. Relative to press statements, which mustappeal to multiple stakeholders and can be altered or coloredby media interpretation, LTIP explanations in proxy

303/ASQ, June 1995

statements speak primarily to shareholders and are notvulnerable to media intervention.

Second, we have identified two specific institutional goals orlegitimate purposes that may be used to explain a variety ofgovernance arrangements and other organizational policies tostakeholders that are not easily classified according to extanttypologies of organizational accounts (cf,, Elsbach, 1994),Whereas the existing literature has focused almostexclusively on reactive forms of impression management,such as denying responsibility for some outcome oraccepting responsibility while attempting to changestakeholders' perceptions about it (e,g,, Sutton and Callahan,1987; Dutton and Dukerich, 1991; Marcus and Goodman,1991; Elsbach and Sutton, 1992), LTIP explanationsrepresent a proactive kind of symbolic management. BothHR explanations and agency explanations serve to preemptor stave off potential criticism of CEO compensation, ratherthan counteract existing criticism. Typologies oforganizational impression management might be enriched bydistinguishing between reactive and proactive explanations.

Finally, we examine explicitly the role of power ingovernance communication. While prior research in this areahas typically assumed that corporate officers and directorscommunicate with stakeholders as a unified group withcommon interests (e,g,, Staw, McKechnie, and Puffer, 1983;Salancik and Meindl, 1984), in this study we acknowledgethe potential for parochial interests among corporate leadersand empirically examine how those divergent interests caninfluence the content of explanations for controversialorganizational action.

It is also noteworthy that we found no support for the notionthat variation in the choice of explanations for new LTIPsmay simply reflect differences in how the plans are actuallyused. Firms using agency explanations did not increasecompensation contingency after LTIP adoption any morethan did those firms not using agency explanations, andfirms using HR explanations did not increase the totalcompensation level after LTIP adoption any more than didthose firms not using such logics. This finding suggests thatLTIPs are not actually used differently by firms usingdifferent explanations and, consequently, that the choice ofexplanations may be driven by symbolic considerations.Since prior symbolic management studies have typically notexplored whether different explanations or justifications fitthe facts of the situation (e,g,, Sutton and Callahan, 1987;Marcus and Goodman, 1991), these supplementary findingsmay also contribute to the symbolic management literature.

This study highlights how a symbolic managementperspective can enhance our knowledge about CEOcompensation issues beyond the traditional focus oneconomic and political perspectives. It is noteworthy thateven within existing CEO compensation debates, thearguments are often couched in symbolic terms,emphasizing whether CEO compensation is justifiable in the

304/ASQ, June 1995

CEO Compensation

eyes of an organization's constituents. We treat CEOcompensation as a justification problem that is sociallydefined, inherently subjective, and also potentially subject tomanipulation. The supportive findings of this study suggestthat our interdisciplinary framework can lead to a greaterunderstanding of the social definition and possiblemanipulation of the phenomenon of CEO compensation. Wehope this framework can also be used to shed light on themany other organizational phenomena that are driven by theinterests of powerful organizational elites but are alsoembedded in an organizational, social, and historical context.

REFERENCES

Alchian, Armen A,, and HaroldDennsetz1972 "Production, infornnation

costs, and economicorganization," AmericanEconomic Review, 62:777-795,

Aldrich, John H,, and Forrest D,Nelson1984 Linear Probability, Logit, and

Probit Models, Newbury Park,CA: Sage,

Allison, Paul D,1984 Event History Analysis,

Newbury Park, CA: Sage,

Arendt, Hannah1986 "Communicative power," In

Steven Lukes (ed,). Power:59-74, New York: New YorkUniversity Press,

Bantel, Karen A., and Susan E,Jackson1989 "Top management and

innovations in banking: Doesthe composition of the topteam make a difference?"Strategic ManagementJournal, 10: 107-124,

Baysinger, Barry, and Robert E,Hosklsson1990 "The composition of boards of

directors and strategic control:Effects on corporatestrategy," Academy ofManagement Review, 15:72-87,

Beatty, Randolph P., and EdwardJ, Zajac1994 "Top management incentives,

monitoring, and risk sharing:A study of executivecompensation, ownership, andboard structure in initial publicofferings," AdministrativeScience Quarterly, 39:313-335,

Berle, Adolph, and Gardiner C,Means1932 The Modem Corporation and

Private Property, New York:Macmillan,

Bettman, James R,, and Barton A.Weitz1983 "Attributions in the board

room: Causal reasoning incorporate annual reports,"Administrative ScienceQuarterly, 29: 57-88,

Blau, Peter M,1977 Inequality and Heterogeneity,

Glencoe, IL: Free Press,

Bus/ness Week1992 "Executive pay," March 30:

52-58,

Byrne, Donn1971 The Attraction Paradigm, New

York: Academic Press,

Chaganti, Rajeswararao, andRakesh Sambharya1987 "Strategic orientation and

characteristics of uppermanagement," StrategicManagement Journal, 8:393^01,

Collins, Randall1979 The Credential Society, New

York: Academic Press,

Crystal, Graef S,1984 Questions and Answers on

Executive Compensation,Englewood Cliffs, NJ:Prentice-Hall,

1991 In Search of Excess: TheQvercompensation ofAmerican Executives, NewYork: W, W, Norton,

Davis, Gerald F,, and Suzanne K,Stout1992 "Qrganization theory and the

market for corporate control:A dynamic analysis of thecharacteristics of largetakeover targets, 1980-1990,"Administrative ScienceQuarterly, 37: 605-633,

Davis, Gerald F,, and Tracy A,Thompson1994 "A social movement

perspective on corporatecontrol," AdministrativeScience Quarterly, 39:141-173,

DeAngelo, Harry, and LindaDeAngelo1991 "Union negotiations and

corporate policy: A study oflabor concessions in thedomestic steel industry duringthe 1980s," Journal ofFinancial Economics, 30:3-43,

Dutton, Jane E,, and Janet M,Dukerich1991 "Keeping an eye on the

mirror: Image and identity inorganizational adaptation,"Academy of ManagementJournal, 34: 517-554,

Eisenhardt, Kathleen M,1989 "Agency theory: An

assessment and review,"Academy of ManagementJournal, 14: 57-74,

Elsbach, Kimberly D,1994 "Managing organizational

legitimacy in the Californiacattle industry: Theconstruction and effectivenessof verbal accounts,"Administrative ScienceQuarterly, 39: 57-88,

Elsbach, Kimberly D,, and RobertI, Sutton1992 "Acquiring organizational

legitimacy through illegitimateactions: A marriage ofinstitutional and impressionmanagement theories,"Academy of ManagementJournal, 35: 699-738,

Finkelstein, Sydney1992 "Power in top management

teams: Dimensions,measurement, and validation,"Academy of ManagementJournal, 35: 505-538,

Finkelstein, Sydney, and RichardA, D'Aveni1994 "CEQ duality as a

double-edged sword: Howboards of directors balanceentrenchment avoidance andunity of command," Academyof Management Joumal, 37:1079-1108,

305/ASQ, June 1995

Finkelstein, Sydney, and DonaldC. Hambrick1989 "Chief executive

compensation: A study of ttieintersection of markets andpolitical processes," StrategicManagement Joumal, 10;121-134,

Foulkes, Fred K.1991 "Introduction," In Fred K,

Foulkes (ed,). ExecutiveCompensation: a StrategicGuide for the 1990s: xvii-xxiv,Boston: Harvard BusinessSchool Press,

Fredrickson, James W,, Donald C.Hambrick, and S. Baumrin1988 "A model of CEO dismissal,"

Academy of ManagementReview. 13: 255-270,

Friedland, Roger, and Robert R.Alford1991 "Bringing society back in:

Symbols, practices, andinstitutional contradictions," InWalter W, Powell and Paul J,DiMaggio (eds,), The NewInstitutionalism inOrganizational Analysis:232-263, Chicago: Universityof Chicago Press,

Gomez-Mejia, Luis R., andTheresa M. Welbourne1988 "Compensation strategy: An

overview and future steps,"Human Resource Planning,11(3): 173-189,

Greene, William H.1993 Econometric Analysis, New

York: Macmillan,

Hambrick, Donald C, and PhyllisP. Mason1984 "Upper echelons: The

organization as a reflection ofits top managers," Academyof Management Review. 2:193-206,

Harrison, J. Richard, David LTorres, and Sal Kukalis1988 "The changing of the guard:

Turnover and structuralchange in thetop-management positions,"Administrative ScienceOuarterly, 33: 211-232,

Hill, Charles W., and Philip P.Phan1991 "CEO tenure as a determinant

of CEO pay," Academy ofManagement Joumal. 34:707-717,

Hitt, Michael A., and Beverly B.Tyler1991 "Strategic decision models:

Integrating differentperspectives," StrategicManagement Journal. 12:327-351,

Holsti, Ole R.1968 "Content analysis," In

Gardner Lindzey and ElliotAronson (eds,), The Handbookof Social Psychology. 2:596-692, Reading. MA:Addison-Wesley,

Homans, George C.1950 The Human Group, New York:

Harcourt. Brace,

Jarrell, Gregg A.1993 "An overview of the executive

compensation debate,"Journal of Applied CorporateFinance. 5: 76-82,

Jensen, Michael C, and WilliamH. Meckling1976 "TheorY of the firm:

Managerial behavior, agencycosts, and ownershipstructure," Journal of FinancialEconomics. 3: 305-350,

Jensen, Michael C, and Kevin J.Murphy1990 "Performance pay and

top-management incentives,"Joumal of Political Economy.98: 225-263,

Jensen, Michael C, and RichardS. Ruback1983 "The market for corporate

control," Journal of FinancialEconomics. 11: 5-50,

Judge, Timothy A., and Gerald R.Ferris1993 "Social context of

performance evaluationdecisions," Academy ofManagement Journal, 36:80-105,

Kanter, Rosabeth M.1977 Men and Women of the

Corporation, New York: BasicBooks,

Kerr, Jeffrey L., and Leslie Kren1992 "Effect of relative decision

monitoring on chief executiveofficer compensation,"Academy of ManagementJournal, 35: 370-397,

Kramer, Roderick M.1991 "Intergroup relations and

organizational dilemmas: Therole of categorizationprocesses," In L, L,Cummings and Barry M, Staw(eds,). Research inOrganizational Behavior. 13:191-228, Greenwich, CT: JAIPress,

Latham, Gary P., Kenneth N.Wexley, and Elliott D. Pursell1975 "Training managers to

minimize rating errors in theobservation of behavior,"Journal of Applied Psychology.60: 550-555,

Lawler, Edward E.1968 "Equity theory as a predictor

of productivity and workquality," PsychologicalBulletin, 70: 596-610,

1990 Strategic Pay: AligningOrganizational Strategies andPay Systems, San Francisco:Jossey-Bass,

Lazear, Edward, and SherwinRosen1981 "Rank-order tournaments as

optimum labor contracts,"Journal of Political Economy.89: 841-864,

Mace, Myles L.1971 Directors: Myth and Reality,

Boston: Harvard BusinessSchool Press,

Marcus, Alfred A,, and Robert S.Goodman1991 "Victims and shareholders:

The dilemmas of presentingcorporate policy during acrisis," Academy ofManagement Journal, 34:281-305,

Marris, Robert L.1964 The Economic Theory of

Managerial Capitalism,London: MacMillan,

Mayo, Elton1945 The Social Problems of an

Industrial Civilization,Cambridge. MA: HarvardUniversity Press,

Meindl, James R., Sanford B.Ehrlich, and Janet M. Dukerich1985 "The romance of leadership,"

Administrative ScienceOuarterly. 30: 78-102,

Messick, David M., and Dianne M.Mackie1989 "Intergroup relations," In

Mark R, Rosenzweig andLyman W, Porter (eds,).Annual Review of Psychology,40: 45-81, Palo Alto. CA:Annual Reviews,

Meyer, John W., and Brian Rowan1977 "Institutional organizations:

Formal structure as myth andceremony," American Joumalof Sociology, 83: 340-363,

Miles, Raymond E., and Charles C.Snow1978 Organizational Strategy,

Structure and Process, NewYork: McGraw-Hill,

Milkovich, George T., and JerryM. Newman1984 Compensation, Piano, TX:

Business Publications,

306/ASQ, June 1995

CEO Compensation

Mizruchi, Mark S.1983 "Who controls whom? An

examination of the relationbetween management andboards of directors in largecorporations." Academy ofManagement Review, 8:426-^35.

Mobley, William H.1982 "Supervisor and employee

race and sex effects onperformance appraisals: Afield study of adverse impactand generalizability." Academyof Management Journal, 25:598-606.

Morrill, Calvin1991 "Conflict management, honor,

and organizational change."American Journal ofSociology, 97: 585-621.

Murphy, Kevin J.1986 "Top executives are worth

every nickel they get."Harvard Business Review,64(2): 125-132.

New York Times1992 "Executives are fuming over

data on their pay." August 25:D1-D3.

O'Reilly, Charles A., Ill, David F.Caldwell, and William P. Barnett1989 "Work group demography,

social integration, andturnover." AdministrativeScience Quarterly, 34: 21-37.

Ouchi, William G.1980 "Markets, bureaucracies and

clans." Administrative ScienceQuarterly, 25: 129-141.

Patton, Arch, and John C. Baker1987 "Why won't directors rock the

boat?" Harvard BusinessReview, 65(6): 10-18.

Perrow, Charles1986 Complex Organizations: A

Critical Essay. New York:Random House.

Pettigrew, Andrew M.1985 "Examining change in the

long-term context of cultureand politics." In J. M.Pennings (ed.). Perspectiveson Organizational Strategy andChange: 269-318. SanFrancisco: Jossey-Bass.

Pfeffer, Jeffrey1981 "Management as symbolic

action: The creation andmaintenance of organizationalparadigms." In L. L.Cummings and Barry M. Staw(eds.). Research inOrganizational Behavior, 3:1-52. Greenwich, CT: JAIPress.

1994 Competitive Advantagethrough People. Boston:Harvard Business SchoolPress.

Rechner, Paula L, and Daniel R.Dalton1991 "CEO duality and

organizational performance: Alongitudinal analysis."Strategic ManagementJournal, 12: 155-160.

Ross, Steven1973 "The economic theory of

agency: The principal'sproblem." American EconomicReview, 63: 134-139.

Salancik, Gerald R., and James R.Meindl1984 "Corporate attributions as

strategic illusions ofmanagement control."Administrative ScienceQuarterly, 29: 238-254.

Singh, Harbir, and Farid Harianto1989 "Management-board

relationships, takeover risk,and the adoption of goldenparachutes." Academy ofManagement Journal. 32:7-24.

Stangor, Charles, Laure Lynch,Changming Duan, and Beth Glass1992 "Categorization of individuals

on the basis of multiple socialfeatures." Journal ofPersonality and SocialPsychology, 62: 207-218.

Staw, Barry M., Pamela I.McKechnie, and Sheila M. Puffer1983 "The justification of

organizational performance."Administrative ScienceQuarterly, 28: 582-600.

Sutton, Robert I., and Anita L.Caliahan1987 "The stigma of bankruptcy:

Spoiled organizational imageand its management."Academy of ManagementJournal, 30: 405-436.

Tajfel, Henri, and John C. Turner1986 "The social identity theory in

intergroup behavior." In S.Worchel and W. G. Austin(eds.). Psychology ofIntergroup Relations: 7-24.Chicago: Nelson-Hall.

Time1993 "Rolling back executive pay."

March 1: 49-50.

Tsui, Anne S., Terri D. Egan, andCharles A. O'Reilly1992 "Being different: Relational

demography andorganizational attachment."Administrative ScienceQuarterly, 37: 549-579.

Tsui, Anne S., and Charles A.O'Reilly1989 "Beyond simple demographic

effects: The importance ofrelational demography insuperior-subordinate dyads."Academy of ManagementJournal, 32: 402-423.

Useem, Michael, and JeromeKarabel1986 "Pathways to top corporate

management." AmericanSociological Review. 51:184-200.

Wade, James B., Charles A.O'Reilly, and Ike Chandratat1990 "Golden parachutes: CEOs

and the exercise of socialinfluence." AdministrativeScience Quarterly, 35:587-603.

Weber, Robert P.1985 Basic Content Analysis.

Beverly Hills, CA: Sage.

Westphal, James, and Edward J.Zajac1994 "Substance and symbolism in

CEOs' long-term incentiveplans." Administrative ScienceQuarterly, 39: 367-390.

Wiersema, Margarethe F., andKaren A. Bantel1992 "Top management team

demography and corporatestrategic change." Academyof Management Journal, 35:91-121.

Williamson, Oliver E.1964 The Economics of

Discretionary Behavior:Managerial Objectives in aTheory of the Firm.Englewood Cliffs, NJ:Prentice-Hall.

Wright, Patrick M., Gary C.McMahan, and AbigailMcWilliams1994 "Human resources and

sustained competitiveadvantage: A resource-basedperspective." InternationalJournal of Human ResourceManagement, 5: 301-326.

307/ASQ, June 1995

Zajac, Edward J. ZaId, Mayer N. Zucker, Lynne G.1990 "CEO selection, succession, 1969 "The power and functions of 1983 "Organizations as

compensation, and firm boards of directors: A institutions." In Samuel B.performance: A theoretical theoretical synthesis." Bacharach (ed.). Research inintegration and empirical American Journal of the Sociology ofanalysis." Strategic Sociology, 75: 97-111. Organizations, 2: 1-42.Management Journal, 11: Greenwich, CT: JAI Press.217-230.

308/ASO, June 1995