when do leaders matter? ownership, governance and the influence of ceos on firm performance

The Leadership Quarterly 25 (2014) 358–372

Contents lists available at ScienceDirect

The Leadership Quarterly

j ourna l homepage: www.e lsev ie r .com/ locate / leaqua

When do leaders matter? Ownership, governance and theinfluence of CEOs on firm performance

Jonathan R. Clark a,⁎,1, Chad Murphy b,2, Sara J. Singer c,3

a The Pennsylvania State University, 118B Keller Building, University Park, PA 16802, USAb The Pennsylvania State University, 439A Business Building, University Park, PA 16802, USAc Harvard School of Public Health, 317 Kresge Building, 677 Huntington Avenue, Boston, MA 02115, USA

a r t i c l e i n f o

⁎ Corresponding author at: Department of Health PolE-mail addresses: [email protected] (J.R. Clark), cbm1

1 Tel.: +1 814 863 2902.2 Tel.: +1 814 863 2384.3 Tel.: +1 617 432 7139.

1048-9843/$ – see front matter © 2013 Elsevier Inc. Ahttp://dx.doi.org/10.1016/j.leaqua.2013.09.004

a b s t r a c t

Article history:Received 14 September 2012Received in revised form 22 July 2013Accepted 16 September 2013Available online 14 October 2013

Editor: John Antonakisis, who owns and monitors the organization. In this paper, we outline how different ownership

Leadership and strategic management research suggests that the extent to which CEOsinfluence performance largely depends on the presence or absence of certain factors. Thesefactors may include the characteristics of the task at hand, subordinates, the organization itselfor the external environment. Among these factors, a fundamental contingency that hasreceived little empirical attention is an organization's ownership and governance structure—that

and governance structures can present the opportunity for, or limit, leader influence andempirically examine the extent to which CEO effects on financial performance depend on thesestructures. Examining organizations in the same industry but with different ownership andgovernance structures, our results suggest that these structures are closely aligned with thedegree to which CEOs influence firm performance. Our findings support the notion that leadersmatter most when ownership and governance structures correspond with a weak orambiguous institutional logic. This study contributes new insight into the “opportunitystructure” of CEO influence, that is, the organizational factors that shape leader discretion and,hence, condition the CEO's level of influence over firm performance.

© 2013 Elsevier Inc. All rights reserved.

Keywords:CEO discretionInstitutional logicsFirm performanceOwnership and governanceOpportunity structure

1. Introduction

Under what conditions are leaders most able to affect their organizations? This question has a long history in organizationalresearch, some of the most vibrant streams about which have come out of the literatures on contingency theories of leadership(Fiedler, 1967; House, 1971; Kerr & Jermier, 1978) and the upper echelons perspective on strategic management (Finkelstein &Hambrick, 1990; Hambrick & Mason, 1984; Peteraf & Reed, 2007). While the focus of contingency theories has been onmicro-level factors that affect leadership effectiveness (Klein, Ziegert, Knight, & Xiao, 2006; Nübold, Muck, & Maier, 2012), theliterature on strategic management has centered on macro-environmental concerns (Crossland & Hambrick, 2007; Hambrick &Abrahamson, 1995). Relatively scant attention has been given, however, to organizational features as a contingent condition forleaders. We seek to address this omission, showing in this study that ownership and governance structures–in particular, theclarity of the associated institutional logic as determined by the degree to which these functions are aligned in a singlestakeholder group–can serve as key influences on leader discretion and, hence, leader effects on firm performance.

Many prior studies have illuminated important factors shaping leader effectiveness. Leadership scholars have primarilyfocused on the role of task characteristics, intrapersonal factors of both leaders and followers, and interpersonal dynamics

icy and Administration, The Pennsylvania State University, 118B Keller Building, University Park, PA [email protected] (C. Murphy), [email protected] (S.J. Singer).

ll rights reserved.

http://crossmark.crossref.org/dialog/?doi=10.1016/j.leaqua.2013.09.004&domain=pdf

http://dx.doi.org/10.1016/j.leaqua.2013.09.004

mailto:[email protected]



http://dx.doi.org/10.1016/j.leaqua.2013.09.004

http://www.sciencedirect.com/science/journal/10489843

359J.R. Clark et al. / The Leadership Quarterly 25 (2014) 358–372

between the two in determining the impact of leaders. For example, path–goal theory, proposed by Robert House (1971, 1996),posits that leader effects are contingent on the alignment of leadership style with various aspects of the task at hand, such as itslevel of ambiguity, and employee characteristics, including capabilities and locus of control. Hundreds of empirical studies havebeen conducted examining contingency theories of leadership, and interest in the moderators and possible substitutes forleadership remains high (Dionne, Yammarino, Howell, & Villa, 2005), with most of this work continuing to center on followers(Dvir & Shamir, 2003; Kellerman, 2008; Nübold et al., 2012; Shamir & Howell, 2000) and the characteristics of the task orsituation (Herrmann & Felfe, 2013; Purvanova & Bono, 2009).

In a similar way, strategic management scholars have pursued a conditional approach to understanding the impact of leaderson their organizations, specifically in terms of the influence of CEOs on firm performance—termed the “leader effect” or “CEOeffect.” However, the focus here has been on macro-environmental concerns, such as regulatory obstacles (Peteraf & Reed, 2007)and industry conditions (Hambrick & Abrahamson, 1995) and how such factors shape the potential effects of leaders. Forexample, the upper-echelons perspective (Hambrick & Mason, 1984) assumes that organizations are a reflection of their topmanagers and specifically top managers' behavioral tendencies as shaped by personal background and other individualcharacteristics (Finkelstein & Hambrick, 1990). Refinements to the theory have specified that top managers can influence a firm'sstrategy and performance only if they have a sufficient degree of discretion for action (Hambrick & Finkelstein, 1987).Organizational theorists have likewise proposed macro-level conditions that pose a threat to the very notion of leader effects,such as industry norms regarding “legitimate” behavior (DiMaggio & Powell, 1983) that induce mimicry among organizations,and evolutionary life-cycles of firm populations that unleash forces beyond the control of any individual leader (Hannan &Freeman, 1977). Situations conducive to leader effects have not been entirely neglected, however. For instance, Salancik andPfeffer (1977) have shown that, in the public sector, administrators can have a greater effect on outcomes to the extent that theyare free from the constraints of powerful parties. In sum, macro-approaches to leader effects consider the “strength” of situations(Mischel, 1977). Weak situations, that is, situations facing weak constraints, offer high levels of managerial discretion and allowleaders to influence organizational outcomes (Barrick & Mount, 1993; House, Spangler, & Woycke, 1991; Jones & Olken, 2005;Tsui, Zhang, Wang, Xin, & Wu, 2006; Waldman, Ramirez, House, & Puranam, 2001).

Despite these contributions to understanding leader effects, comparatively little attention, in either research tradition, hasbeen paid to contingency factors at the organizational level. To be sure, scholars have recognized the important influence offactors such as organizational identity on organizational action (Dutton & Dukerich, 1991; Dutton & Penner, 1993; Eccles &Nohria, 1992). Yet these factors have largely been left unexamined empirically, particularly when it comes to the influence ofCEOs on the performance of their organizations Hambrick and Finkelstein (1987). Note that this mid-level of analysis–includingorganizational structure and process–is likely an important source of moderators of CEO effects. Similarly, Shamir and Howell(1999) theorize that organizational context–including technology, goals, structure, culture and governance–may shape theeffectiveness of charismatic CEOs depending on the strength of the situation they face. Our paper addresses this gap by focusingon the role of ownership and governance in the degree to which CEOs matter for firm performance.

We consider ownership and governance for two reasons. First, as agents of the firm, CEOs may be the most directly freed orlimited by these organizational structures. According to a recent review of the influence CEOs exert on strategic change, the mosttypical source of CEO constraint may be the direct “mandate” provided by boards of directors or owners to either stay the courseor change strategic direction (Hutzschenreuter, Kleindienst, & Greger, 2012). Second, ownership and governance structures maybe closely associated with specific institutional logics (e.g., for-profit ownership is associated with a market-oriented logic) thatcan powerfully yet often indirectly open or limit CEOs' opportunities to exercise discretion (Thornton, 2002; Thornton & Ocasio,2008). As Thornton and Ocasio (1999) have noted, “institutional logics define the rules of the game by which executive power isgained, maintained, and lost in organizations” (p. 802). Prior studies have linked CEOs to ownership and governance in thecontext of certain outcomes, such as executive compensation (Hambrick & Finkelstein, 1987). Goodstein and Boeker (1991) alsolinked governance to CEOs and strategic change, finding that ownership and governance–specifically, which groups are incontrol–can strongly impact a CEO's freedom to determine strategic direction.

While the latter study emphasizes the influence of ownership and governance on a CEO's freedom to determine a course ofaction, ownership and governance have not yet been examined empirically with respect to a CEO's effect on firm performance.Thus our study is motivated by the following research question: under what types of ownership and governance arrangementswill a CEO have the greatest effect on the financial performance of his or her firm?We build on prior research to show empiricallythat CEO effects on organizational performance are contingent upon ownership and governance arrangements, and we advancetheory by showing that such contingent effects depend largely on the clarity or ambiguity of the associated institutional logic. Indoing so, we offer both empirical and theoretical insights into the question of leader effects on firm performance. While themechanisms underlying CEO effects are beyond the scope of this study, we seek to explain the “opportunity structure” for sucheffects (McAdam, 1996), that is, the conditions external to the CEO that determine his or her “window of opportunity” forinfluencing firm performance.

2. Theory and hypotheses

Before developing our hypotheses, we should define more precisely “leader effect” or “CEO effect.” In the stream of research onstrategic leadership, the “leader effect” is “the proportion of variance in a firm-level outcome variable that is statisticallyassociated with, or can be attributed to, the presence of individual CEOs in the sample” (Crossland & Hambrick, 2007: 769–770).We prefer the term “CEO effect” to recognize that leadership may be exercised without regard to role. We also refer to CEO rather

360 J.R. Clark et al. / The Leadership Quarterly 25 (2014) 358–372

than top management teams to reflect more accurately our data. We acknowledge, however, that teams may be the moreappropriate locus of leadership in hospitals. Nevertheless, since top managers often turn over with CEO succession, we regardthese as similar constructs. Our approach, which we describe more fully in the Section 3, is designed to determine the extent towhich discrete CEOs explain performance beyond what contextual factors (e.g., macro-economy, organization-level factors)would predict (Crossland & Hambrick, 2007).

In this section we propose hypotheses that predict the ownership and governance conditions under which CEO effects arelikely to be the strongest. Following prior studies on CEO effects on performance, we conceptualize our hypotheses in a cross-levelcomparative fashion. That is, we theorize about the relative strength of CEO, organizational, and external effects on performanceunder different ownership and governance arrangements. Such a multi-level approach has a long tradition in strategicmanagement, with studies tending to find a CEO effect on financial performance somewhere between 5% and 20% (Bertrand &Schoar, 2003; Crossland & Hambrick, 2007; Thomas, 1988; Weiner, 1978), some even suggesting up to 45% (Day & Lord, 1988). Indeveloping our hypotheses, our use of the term “external environment” refers broadly to the institutional environment in whichan organization is situated. More specifically, “external” refers to the market conditions for the organization in any particular year,which includes regulatory standards, political interests, regional socioeconomic conditions, and other geographically proximatecompetitors.

Our hypotheses link the governance and ownership structure of a firm with performance effects through institutional logics(Friedland & Alford, 1991). An institutional logic is defined as “the socially constructed, historical patterns of material practices,assumptions, values, beliefs, and rules” that create norms around “the way a particular social world works” (Jackall, 1988: 112;Thornton & Ocasio, 1999: 804). For example, an organization may operate in accordance with a market-oriented institutionallogic, wherein shareholder returns are of utmost importance, or a professional-oriented institutional logic, wherein theorganization is expected to function largely through peer-to-peer relationships instead of hierarchy (Thornton, 2002). Suchexpectations “provide both a logic of action and [reinforce] a set of cultural and material values” for the organization (Ocasio,1997: 196). Institutional logics are often viewed as originating from a “supraorganizational” source, such as from withininstitutional sectors that span societies (e.g., state bureaucracy) (Thornton & Ocasio, 2008: 101). However, an organization'sdominant logic may be most tangibly reflected in various organization-level factors, such as organizational missions (Thornton,2002; Thornton & Ocasio, 2008).

Institutional logics are also reflected in ownership and governance arrangements: for-profit organizations–owned andgoverned by shareholders–tend to prioritize shareholder returns, thus reflecting a market-oriented logic, while non-profitorganizations–owned and governed by the community–tend to place more value on service to the community, indicating acommunity-oriented logic. These logics thus capture the interests and values of those who own and govern the organization andspecify the stakeholders (e.g., politicians, shareholder, the community) who should be deemed most important. Indeed, logics arenot pure abstractions—they concretely affect day-to-day affairs of an organization. A particular logic determines not only “themeaning and legitimacy” (Thornton, 2002: 83) of the organization's attributes, but also it “focuses the attention of organizationalactors on a limited set of issues and solutions that are consistent” with it (Thornton, 2002: 83), thereby placing bounds on thelatitude of decision-makers. Thus, to the extent that ownership and governance affects the impact of CEOs, the underlying force ofthis effect may derive, in part, from the associated institutional logic, specifically the normative expectations it creates regardinghow the organization should conduct its affairs and the objectives toward which it should strive. These expectations have farreaching implications, and may even influence power dynamics in and around organizations, such as the direct constraints thatowners may place on organizations. As Thornton and Ocasio (1999) have noted, the source of “power, its meaning and itsconsequences are contingent on [the relevant] higher-order institutional logics” (Thornton & Ocasio, 1999, pg. 802). Nevertheless,because normative expectations are indirect, values-based beliefs regarding how people and organizations should behave,institutional logics may affect the opportunity for CEOs to impact performance independent of owner-board-CEO powerdynamics. In other words, working through socio-cognitive channels (rather than lending meaning and force to a directdemonstration of power), institutional logics can constrain the CEO's perceived range of action by simply solidifying in the mindsof organizational members the norms and values the organization should be upholding, which necessarily restricts the CEO'sdecisions to only those considered “acceptable”. In this paper, we seek to complement prior research by focusing on the role ofsuch logic-based constraints in limiting CEO discretion.

Consider the case of publicly owned organizations directly controlled by the government (hereafter “public-direct”organizations). In these organizations, political leaders may directly impose a central purpose or set of purposes on theorganization for political expediency (Cuervo & Villalonga, 2000). Public agencies created by politicians or interest groups aretypically controlled via governance mechanisms, such as “legislative oversight, presidential management, control overappointments,” designed to ensure that the agency's actions serve the objectives of the government by which it is controlled(Moe, 1990: 122). For this reason, public-direct organizations have a tendency to be oriented toward the external interests ofpoliticians, bureaucrats and political cycles. The associated politically-oriented logic makes these organizations subject to thewhims of public choice (Buchanan, 1972; Cuervo & Villalonga, 2000). Politicians may thus impose guidelines regarding theallocation of organizational resources, which limit the leader's ability to shape the strategic orientation of the organization andthereby leave a mark on the firm's performance (Salancik & Pfeffer, 1977). While politically-adept leaders may be able to sidestepcertain elements of direct control, the realities of the ownership and governance structure itself (e.g., legislative oversight, controlover appointments) are likely to be far less avoidable. In a subtler manner, institutional logics may filter through the entireorganization, limiting the willingness of organizational members, including leaders, to accept, consider, or even conceive ofactions that fall outside of the “rules” specified by the dominant logic. For instance, given that financial concerns (i.e., returns to


shareholders) are commonly associated with market-oriented logics, the members of politically-oriented organizations areunlikely to emphasize this concern over and above the political or bureaucratic concerns often associated with government.Indeed, one leader of a county-controlled hospital that we spoke with described the road-blocks he faced in trying to improve theorganization's efficiency by saying, “there's a subculture that says, ‘we're the county, and you don't understand, and it takes uslonger, and you have to do this and that.’” Such deeply entrenched assumptions are indicative of a dominant politically-orientedlogic, rooted in the history of the political environment (in this case “the county”). Under such conditions, an organization and itsleaders may have limited choices in the outcomes the organization pursues or even those its leaders can conceive as appropriatepossibilities, leaving their fate subject to the political and regulatory cycles outside of the organization. Where such conditionsprevail, the differential effects of environment, organization, and CEO on firm financial performance may be hypothesized thus:

Hypothesis 1. When ownership and governance structures are indicative of a clear, politically-oriented logic (e.g., ingovernment-owned, directly-governed hospitals), external effects on performance will dominate both CEO and organizationaleffects.

In contrast to the dominance of environmental effects for public-direct organizations with a politically-oriented logic, otherownership and governance arrangements give more autonomy to the organization itself while still denoting very clear normativeexpectations for behavior. For example, while private for-profit and private non-profit organizations serve stakeholder interestsoutside of the organization (i.e., shareholders or the community), these interests are more narrowly defined around theorganization's mission (rather than oriented toward broader political interests) and the outcomes it can produce (i.e., profit in thecase of for-profits; community well-being in the case of non-profits). Thus, a key distinction between these organizations andpublic-direct organizations is that the “rules of the game” are determined by the narrow purpose of the organization rather thanthe political and bureaucratic whims of the government in control. Thus, the market-oriented logic of for-profits is enactedthrough a focus on shareholder returns, and the community-oriented logic of non-profits is enacted through a focus oncommunity well being (or some other high-order, often service-oriented goal). In both cases, the logic is narrowly centered onoutcomes under the organization's control (rather than those based on political and regulatory expediency), and as a result the“patterns of material practices, assumptions, values, beliefs, and rules” (i.e., the logic) (Thornton & Ocasio, 2008: 101) are likely tobe reinforced through organizational (rather than environmental) history and experience. Despite this increase in autonomyafforded to a private organization around how to execute its mission, the logic indirectly guiding it is just as clear and unalterableas in publicly owned organizations, thus diminishing leader influence (Shamir & Howell, 1999). Indeed, while the ownership andgovernance structures of for-profit firms typically permit internal determination of structures, processes, and strategic direction(Cuervo & Villalonga, 2000), these organizations are still subject to the boundaries of an underlying market-oriented logic, tiedto the central purpose of the organization's existence: to maximize shareholder returns. While the details may be idiosyncraticto each organization, for-profit governance systems (e.g., CEO incentive and pay packages) and organizational structures(e.g., corporate hierarchies) are largely arranged so as to achieve this overriding goal, and organizational norms and values willoften develop in ways that reinforce this core objective. In a similar way, the logics of non-profits are unambiguously centered onmaximizing the well being of targeted stakeholder groups. In non-profits, as with for-profits, these logics are deeply integratedinto the structures and mission of the organization. As Glaeser (2003) has pointed out, although many non-profits have “murkymissions” that permit flexible interpretation and action, those that “diverge too far from their mission statement may be subjectto legal challenges” (2,4). Therefore, he argues, most non-profits do not undergo major revisions of their mission at the hands ofworkers or management (Glaeser, 2003: 2).

Thus, while we expect public-direct and private organizations to differ on whether external versus organizational factorsdominate performance, we also expect that the clear, dominant logics among private for-profit and private non-profitorganizations will limit CEO effects in both cases—through governance mechanisms supportive of that logic and throughembedded organizational norms and values, as outlined above. These descriptions do not imply that CEOs would have zero effectson the outcomes of for-profit and non-profit firms (see Bertrand & Schoar, 2003), but rather that they are limited by governanceprocesses and institutional logics deeply rooted in the organization's (rather than the environment's) history and purpose. Hence,we hypothesize that in privately owned firms, the organization itself will account for more influence over firm financialperformance than either environmental or CEO effects:

Hypothesis 2. When ownership and governance structures are indicative of a clear market- or community-oriented logic (e.g., aswith private for-profit and private non-profit organizations), organizational effects on performance will dominate both CEO andexternal effects.

While public-direct, for-profit and non-profit organizations generally operate with unified ownership and governancestructures (i.e., the owners or their agents also directly govern the firm), which exert an unambiguous dominant logic, there isanother category of organizations that operate with mixed ownership and governance arrangements. These arrangements–whichmay reflect contrasting, even contradictory, logics–create the potential for ambiguity, e.g., in situations where the owners or theiragents do not directly govern the firm. While we focus here on ambiguous logics, we acknowledge that ambiguity inorganizations can arise for other reasons, including on account of organizational change (e.g., a shift of core business because of aturbulent industry, a joint venture, or a corporate spin-off) (Corley & Gioia, 2004; Whetten & Godfrey, 1998: 166–167) or becauseof conflict among top decision-makers over interpretations of the core purpose of the organization (Voss, Cable, & Voss, 2006).Other types of “crises” (e.g., a failed venture) likewise may make questions of core organizational purpose more salient and less


readily answerable (Whetten & Godfrey, 1998: 167). Such ambiguity is a sign of a “weak” situation, because it produces multiple,sometimes conflicting interpretations of what is at stake for the organization, which can induce confusion about “who theorganization is” and muddy expectations about appropriate courses of action (Mischel, 1977). Others (Shamir & Howell, 1999)have argued that leaders operating in such “weak” situations are likely to have more leeway, autonomy, and thereforeresponsibility over their behavior than leaders in situations that are high in situational strength. For example, certain hospitals,while technically owned by the government, have governance structures that insulate them from direct government control(i.e., mixed ownership and governance). In theory, such structures provide these publicly-owned, autonomously-governed(hereafter “public-autonomous”) hospitals the freedom to operate more like private ones (e.g., managers are not appointed bypoliticians and board meetings do not have to be open to the public), yet the organization retains its public status, leading to alevel of haziness regarding its mission and values and the institutional logic to which it subscribes (Voss et al., 2006). Under thesecircumstances, where not only is the organization relatively free from direct external control (i.e., power-based sources ofrestraint), but also the logic-driven normative expectations for behavior are ambiguous, a CEO's discretion is likely to becomparatively high given the “weakness” of the situation. This expectation is consistent with the propositions made by Shamirand Howell (1999), who posited that leaders matter more in comparatively weak organizational situations. Accordingly, whenexternal control is relatively low and logics are ambiguous, CEO discretion is high. In such weak situations, we expect that CEOeffects will be more prominent than both organizational and external effects on financial performance. Thus, we predict thefollowing:

Hypothesis 3a. When ownership and governance structures are arranged such that the institutional logic is ambiguous (e.g., as inpublic-autonomous hospitals), CEO effects on performance will dominate external and organizational effects.

Hypothesis 3b. CEO effects on performance will be greater when ownership and governance structures provide an ambiguousinstitutional logic (e.g., as in public-autonomous hospitals) than CEO effects when ownership and governance provide a clearinstitutional logic (i.e., as in public-direct, non-profit and for-profit hospitals).

In summary, our hypotheses propose that CEO effects depend on the alignment of the organization's underlying ownershipand governance structures because of the associated institutional logics. Specifically, where ownership and governance areunified and associated with a clear, dominant or “strong” logic (e.g., as in public-direct, private for-profit and private non-profitorganizations), CEO effects will be dominated by external- or organization-level effects—external effects in the case ofpolitically-oriented public-direct organizations and organization-level effects in the case of market- or community-orientedprivate organizations. In contrast, where ownership and governance are separated (i.e., owners and their agents do not directlygovern the firm) and the associated logic is ambiguous or “weak”, we expect CEO effects to prevail. See Fig. 1 for a summary ofthese hypotheses.

3. Methods

3.1. Overview of research design

There are a number of methods that might be used to assess CEO influence. One approach would be to identify panel-varyingcharacteristics at the CEO level and then use regression analysis to ascertain whether their relationship with financialperformance varied across ownership and governance structures. As noted by previous researchers, while this approach mightreveal the extent to which specific CEO characteristics matter, “it would fall very short of revealing the overall effect of discreteleaders” (Crossland & Hambrick, 2007, p. 779). Because of this limitation, most prior research on CEO effects has insteadattempted to partition variance in performance using fixed effects ANOVA (Bertrand & Schoar, 2003; Crossland & Hambrick, 2007;

Unified Ownership & Governance

Ownership: Government Governance: Government

Politically Oriented Logic

External effects are dominant

Unified Ownership& Governance

Ownership: Community Governance: Community

Community Oriented Logic

Organizational effects are dominant

Market Oriented Logic

Organizational effects are dominant

Separated Ownership & Governance

Ownership: Government Governance: Community

Ambiguous Logic

CEO effects are dominant

Public-Direct Public-Autonomous Non profit For profit

Unified Ownership & Governance

Ownership: Shareholders Governance: Shareholders

Ownership & Governance

Predominant Effects on Performance

Hospital Type

Institutional Logic

Fig. 1. Ownership, governance, institutional logic and performance.


Thomas, 1988; Weiner, 1978) or random effects techniques (Crossland & Hambrick, 2011; Hough, 2006; Misangyi, Elms,Greckhamer, & Lepine, 2006). These methods allow researchers to estimate the significance of entire classes of effects (e.g., CEOeffects, company effects) without having to specify and measure the countless aspects that constitute that class. The intuitionbehind these techniques has been described by Crossland and Hambrick (2007) as follows:

Consider a data set that reports the annual return on assets (ROA) for each of 100 firms (20 firms from each of five industries)over 15 years. This 1500-cell matrix has a grand mean of x and a grand variance of s2. Some of the variance can be explained bycontextual conditions. For example, each calendar year may explain some variance, as a reflection of the macro-economicconditions. Industry membership will explain more variance, as some industries are systematically more (or less) profitable thanothers. And there will be some ‘firm effects’, as a reflection of the persistent differences in resource endowments, scale, and otherrelatively stable attributes of the individual companies in the sample …. The question then arises as to whether additionalvariance is traceable to the CEOs who are in place for each firm-year. Each firm-year, then, is assigned to the CEO in place thatyear, akin to… creating a dummy variable for each CEO in the sample. A “CEO effect”will be observed to the extent that individualCEOs have distinctive and persistent patterns of performance during their tenures. Namely, for CEOs to ‘matter’ in a statisticalsense, they must deliver performance that diverges from what contextual conditions would predict …. If CEOs areinterchangeable, completely imitative, or highly constrained in their actions, then the CEO effect … will be correspondinglylow. (770).

Following this approach, the test of our hypotheses proceeds in two distinct phases, described in more detail below. We firstseparately estimate leader effects by governance type. That is, we estimate the proportion (%) of variance in performance that isattributable to different classes of effects (including CEO effects) for organizations under different ownership and governancearrangements. Second, following the method used by Crossland and Hambrick (2007, 2011), we statistically compare theseestimates to test for significant differences across ownership/governance types. This latter test is aimed at determining the extentto which CEO effects depend on governance arrangements (e.g., whether CEO effects for public-direct organizations are smallerthan CEO effects for public-autonomous organizations).

3.2. Data

We perform our analysis using data on the financial performance of California hospitals from the California Office of StatewideHealth Planning and Development (OSHPD). OSHPD collects and audits financial statements from each hospital in the state on anannual basis and makes the data publicly available. Thus, the unit of observation in the data is the hospital-year. OSHPD financialdata contains information typically found on income statements, balance sheets and cash flow statements, among other dataelements. We obtained 10 years of data, covering the period between 2001 and 2010. There were 503 total hospitals included inthe dataset in 2001 and only 441 in 2010. The change is primarily due to hospital closures and consolidations. As described inmore detail below, we also identified markets, CEOs, systems, and ownership/governance categories based on the AmericanHospital Association's (AHA) annual survey of hospitals. We note that the AHA data were only available for 370 of the 503 totalhospitals in the OSHPD data set.

3.3. Identifying hospital CEOs

We relied on two data sources to identify hospital CEOs: (1) the OSHPD data itself, and (2) the American Hospital Association's(AHA) Annual Survey of hospitals. Both data sources include the name of each hospital's CEO at the time of data submission foreach year in question. We relied on the OSHPD data as our primary source of information on hospital CEOs and used the AHAAnnual Survey for confirmation and to supplement vague or ambiguous information. For example, a small number of hospitalsidentified the health system CEO as the CEO of the hospital in the OSHPD data. Because our interest is in the CEO of each specificfacility, we used the AHA data to identify the CEO of the individual hospital in these few cases.

Following prior studies on CEO effects (e.g., Mackey, 2008), we exclude from our analysis hospitals that did not experience achange in leadership over the study period. This is done to ensure that CEO effects can be distinguished from facility (i.e., hospital)effects. For hospitals that did not change leaders, CEO effects and hospital effects would be indistinguishable. Accordingly, theirinclusion in the model would ascribe all of the variation in performance for those hospitals to the hospital alone, leading to anoverestimate of hospital effects and an underestimate of CEO effects. After these exclusions, and accounting for outliers andmissing data, our final sample includes 776 CEOs in 333 California hospitals covering the period 2001 through 2010. We note thatdue to the exclusion of outliers, missing data and hospital closures and consolidations, this panel of data is unbalanced, with atotal of 3133 observations. Specifically, 70 of the 333 hospitals had fewer than 10 observations.

3.4. Identifying contextual factors

Our analysis estimates CEO effects after accounting for contextual effects. We specifically examine four classes of contextualeffects operating at different levels of analysis: the local market, the health system (i.e., the corporate-parent), the facility (i.e., theorganizational unit itself), and year. We define each class of effects as follows. First, local markets are defined based on the healthreferral regions (HRRs) developed for the Dartmouth Atlas of Health Care by the Dartmouth Institute for Health Care Policy andClinical Practice (Sutherland, Fisher, & Skinner, 2009; Wennberg & Gittelsohn, 1973). HRR information was derived from the AHAAnnual Survey. HRRs are defined based on referral patterns, e.g., where patients from specific service areas actually receive health

Table 1Return on assets by ownership/governance type.

n Mean (%) Std. Dev. Min Max

All ownership types 3133 1.864 17.10 −89.98 63.23Public-direct (ROA) 217 7.403 21.95 −60.96 59.59Public-autonomous (ROA) 381 0.179 12.07 −77.93 56.28Non-profit (ROA) 1489 2.452 13.86 −87.89 63.23For-profit (ROA) 1046 0.534 20.88 −89.98 62.20


services. Their inclusion in the model as local market effects is intended to capture differences in the institutional, competitiveand demographic environments, e.g., the number of hospitals, the concentration of market share or regional socio-economiccharacteristics. While “local markets” have not yet been examined in the corporate strategy literature, we include them herebecause competition among hospitals, like many industries, is more local than industry-wide.

Second, health systems are defined based on the corporate owner of each facility, as indicated by the AHA Annual Survey. Inmany cases the corporate owner is the hospital itself, in which case the hospital is defined as “freestanding” (i.e., not a member ofa multi-hospital system). Accordingly, in our analysis system effects only arise if a hospital is a member of a multi-hospital system(i.e., not freestanding). System effects are similar to corporate-parent effects described in previous studies (e.g., McGahan &Porter, 2002) and are intended to capture differences in corporate-level factors that may influence hospital financial performance,e.g., how resources are managed and shared across facilities. Third, facility effects relate to the hospital itself and are intended tocapture the impact that the characteristics of individual hospitals (e.g., its stable structures and processes) have on performance.Individual hospitals are assumed to be relatively circumscribed businesses, akin to a business-unit in the corporate strategyliterature, and are distinguished in the data by a unique identifier assigned by OSHPD. Finally, year is simply the year in whichhospitals recorded accounting profits, from 2001 to 2010. The inclusion of year effects is expected to capture temporal changes ina hospital's environment, including inflation, regulatory change and other macro-economic fluctuations. The collective inclusionof all these effects in our models is intended to capture the context within which CEOs operate, from macro-economic factors(year effects) to the organization itself (facility effects) and everything in between (local market effects and health systemeffects).

3.5. Classifying governance structures

Hospital governance type was determined using the AHA Annual Survey, which includes a variable indicating the entity thatcontrols each facility. Specifically, the AHA survey assigns hospitals to one of the following categories: (1) State, (2) County,(3) City, (4) City-County, (5) Hospital District or Authority, (6) Church Operated, (7) Other Not-for-profit, (8) Individual,investor-owned, (9) Partnership, investor-owned, and (10) Corporation, investor-owned. Following an approach taken in previousstudies (e.g., Kane, Singer, Clark, Eckloo, & Valentine, 2012), we aggregated these categories as follows: 1―4 (public-direct),5 (public-autonomous), 6―7 (private-non-profit) and 8―10 (private-for-profit). These categories capture key differences inownership and governance. Specifically, public-direct and public-autonomous hospitals are owned by the government, whereasthe others are privately owned entities, non-profits being “owned” by the community (in the sense that the rights to theorganization's residual assets belong to the public at large), and for-profits being owned by shareholders. These organizations alsodiffer in their governance, i.e., who maintains monitoring and decision rights over the organization. Non-profit and for-profitorganizations are generally governed by a board of directors, which serves as the agent of the owners. Public-direct owners oftengovern the organization directly (e.g., direct oversight of the CEO by a county manager) or appoint a board as agent of the owner.Despite the differences, in each of these three cases, the owner or its agent retains the right of governance. However, in the case ofpublic-autonomous hospitals, the owner of the facility (e.g., a public institution) cedes the right of governance to the community.In this case, the organization is publicly owned, but governed by an autonomous board, elected by the community it serves.

3.6. Dependent variable

The dependent variable in our analysis is return on assets (ROA). ROA is calculated as net income (total revenues minus totalexpenses) divided by total assets. Our dataset included several values for ROA that represent extreme outliers and are likely theresult of data errors (e.g., an ROA of 7 or −5). In order to eliminate the possibility of these extreme values exerting undueinfluence on our findings we remove the top and bottom 1% of the distribution from our sample. Table 1 presents summarystatistics for ROA by governance type.4 Fig. 2 presents the trend in average ROA between 2001 and 2010.

We recognize that there are other measures of operating performance that we might use (e.g., total margin). However, theyare highly correlated with ROA (r = 0.81). Moreover, ROA is a measure of accounting profit used commonly in the strategyliterature on performance variation (Crossland & Hambrick, 2007; Mackey, 2008; McGahan & Porter, 2002), as well as in priorstudies on leadership and its direct and indirect effects on firm performance (Avolio, Waldman, & Einstein, 1988; Colbert,

4 We recognize that the averages presented in Table 1 may seem counterintuitive (e.g., public-direct hospitals have the highest average ROA). However, wenote here that the observed differences across categories are consistent with previous research on hospital profitability (Kane et al., 2012).

Fig. 2. Average return on assets (%) by year.


Kristof-Brown, Bradley, & Barrick, 2008). We furthermore recognize that financial performance, on the surface, may not appear tobe the most relevant outcome for hospital CEOs, particularly those under non-profit and government ownership. Certainly, from asocial perspective, we hope that financial considerations never trump the health of our friends and loved ones as a measure ofperformance. We note, however, that regardless of whether hospitals pursue returns for shareholders, in a market economy likethe US, earning a profit sufficient to support continuing operations, replace and invest in capital, and fund growth is no lessimportant for these organizations than it is for for-profit firms. Consistent with this view, we note that a growing body ofliterature supports the notion that non-profit firms in the US health care industry often behave like profit maximizers (Vitaliano,2003). Regardless, hospital CEOs themselves have consistently indicated that financial considerations are their number oneconcern.5 Given this, when it comes to the work and concerns of hospital CEOs in the US, financial returns may be among the mostrelevant outcomes. Moreover, quality measurement of US hospitals is relatively immature. Publicly available measures of hospitalquality are limited (i.e., do not necessarily reflect overall performance) and would not allow for a robust analysis over areasonable period of time. Despite the limitation associated with our chosen outcome measure, we present the results of severalrobustness analyses following the presentation of our base findings.

3.7. Analytical model

Having identified the local market, hospital system, facility, CEO and year associated with each observation in our dataset weexamine the following model of hospital financial performance as measured by ROA:

5 Accache.or

6 Thethese v

rhijkt ¼ μ þ γt þ αh þ βi þ δ j þ φk þ Xjt þ εhijkt : ð1Þ

In this equation rhijkt represents the ROA in year t for CEO k in hospital j, a member of system i, operating in local market h. TheGreek letter μ represents the average ROA over the entire time period; γt represents the premium over the average in year t; αh

the premium associated with being in local market h; φk the premium associated with being a member of system i; δj thepremium associated with hospital j; Xjt the premium associated with CEO k; Xjt represents a vector of hospital variables; and εhijktrepresents the residual. Hospital variables are included to adjust for differences in financial risk and the influence of hospital sizeon performance. These controls are captured based on each hospital's debt-to-equity ratio (i.e., degree of leverage) and thenumber of patients discharged from each facility (i.e., patient volume).6

Following the research design described previously, we examine Eq. (1) using a multi-level random effects approach toestimate the share of variance attributable to each effect class. While previous studies have relied on various fixed effects ANOVAtechniques (e.g., Crossland & Hambrick, 2007; Lieberson & O'Connor, 1972; Mackey, 2008; Weiner, 1978), the multi-level randomeffects approach offers some important advantages. Specifically, while for the most part our data is hierarchically nested (yearsare nested within CEOs, CEOs are nested within hospitals, hospitals are nested within systems and systems are nested withinmarkets), it is possible that some cross nesting exists. For example, while a given hospital may be nested in only one system andone market, it is possible that a given system has two hospitals operating in different markets. In such cases, there may be somedegree of dependence between market and system effects. Moreover, the question of whether system or market resides at thehigher level in the nesting hierarchy creates an “order-of-entry” problem for fixed effects ANOVA (Thomas, 1988). This is because

ording to the American College of Healthcare Executives CEOs have rated financial challenges their chief concern for eight years running. See: http://www.g/PUBS/research/ceoissues.cfm (accessed Dec 12, 2012).se variables can be included in our model because of the random effects estimation approach we use. In a fixed effects model (see our robustness analysis)ariables would be treated as part of the hospital fixed effect.

http://www.ache.org/PUBS/research/ceoissues.cfm

http://www.ache.org/PUBS/research/ceoissues.cfm

Table 2Partitioning of variance in return on assets: multi-level random effects.

Public-autonomous (PA) Public-direct (PD) Non-profit (NP) For-profit (FP)

n 381 217 1489 1046

Variance % Total variance⁎ Variance % Total variance⁎ Variance % Total variance⁎ Variance % Total variance⁎

Year b0.1 b0.01% 188.20 39.07% b0.1 b0.01% 37.80 7.59%Market 4.78 3.13% 141.18 29.31% 8.93 3.77% 26.45 5.31%System b0.1 b0.01% b0.1 b0.01% 10.13 4.27% 77.57 15.58%Facility b0.1 b0.01% b0.1 b0.01% 99.49 41.94% 114.75 23.05%CEO 53.47 35.04% 121.38 25.20% 39.53 16.67% 75.88 15.24%Residual 94.36 61.80% 30.90 6.42% 79.12 33.35% 165.41 33.22%

⁎ Denotes the estimated % of total variance in return on assets attributable to each effect class.


fixed effects ANOVA techniques estimate variance components by sequentially adding each effect class to the model from thehighest level in the hierarchy to the lowest, and noting the incremental change in R-squared. Thus, without perfect nesting, fixedeffects ANOVA techniques may produce equivocal results depending on the order of entry. Multi-level estimation procedurespermit more complicated relationships between effect classes and avoid the order of entry problem by using “iterative estimationto simultaneously estimate all variance components” (Hough, 2006, p. 50).

This iterative procedure is accomplished using restricted maximum likelihood estimation (RMLE) methods,7 which canbe used to directly estimate the portion of the variance in firm performance that is attributable to each random effect class(i.e., market, system, hospital, CEO, year). In order to compare the relative contribution of each effect class, we simply divided theresulting variance components (market, system, facility, leader and year) by the total variance in firm performance. These valuesare directly analogous to the incremental, or partial, R2s from fixed effects models (Crossland & Hambrick, 2007; Mackey, 2008),in that they describe the proportion of the total variance in firm performance that is explained by each effect class.

We ran this procedure for each governance type separately and compared the estimated variance components across types. Inorder to test our hypotheses related to the differences in leader effects across governance types, we convert the proportion ofvariance explained by each effect class into a partial r, following the approach used by Crossland and Hambrick (2007).Specifically, we take the square root of each proportion and use Fisher's z-test for correlation differences to make statisticalcomparisons across governance types. For example, we compare the variance attributable to CEO effects for non-profits withmarket, system, facility and year effects for the same (as a test of hypothesis 2). As a test of the central issue under considerationin this paper, we compare CEO effects for public-autonomous hospitals with CEO effects for public-direct, non-profit andfor-profit hospitals (as a test of hypothesis 3b). Our hypotheses are tested based on the p-value associated with the z-statistic foreach comparison. We note, as have others (Crossland & Hambrick, 2007:780; Bobko, 2001:55), that Fisher's z-test is a rigid test ofsignificance. Thus, any significant differences should be viewed as conservative estimates of actual differences in leader effects.

4. Results

The results of our analysis of Eq. (1) are presented in Table 2. For each governance category, the table reports both the rawvariance component and the corresponding percentage of total variance attributable to each effect class.

With respect to Hypothesis 1, Table 2 indicates that among public-direct hospitals nearly all of the variance is apportioned toyear, market and CEO effects (a minuscule amount is apportioned to each of the other categories). Notably, while our hypothesessuggest a lesser role for CEOs of public-direct hospitals, our findings suggest that such CEOs still influence performance to a greatdegree (25.2% of variance explained). Nevertheless, Hypothesis 1 suggested that, for public-direct hospitals, external effects (yearand market) would dominate CEO and organization-level (facility and system) effects. As Column 2 in Table 2 indicates, thevariance explained by year and market effects is greater than both facility and CEO effects. Individually, year effects aresignificantly different from CEO effects (p b 0.05) though market effects are not (p = 0.288). Nevertheless, in aggregate, externaleffects (year + market = 68.38%) represent nearly three times the effect of CEOs among public-direct hospitals and aresignificantly greater than facility effects (p b 0.01). These estimated year and market effects for public-direct hospitals are alsostatistically greater than the estimated year and market effects for public-autonomous (p b 0.01), non-profit (p b 0.01) andfor-profit (p b 0.01) hospitals. In summary, we find support for Hypothesis 1. However, our findings also suggest that CEOs stillexplain a substantial portion of the variance in performance among public-direct hospitals.

With respect to Hypothesis 2, Table 2 indicates that organization-level effects (e.g., system and facility) are most important forthe performance of hospitals under non-profit and for-profit governance arrangements. Specifically, among the non-profithospitals, facility effects (41.94%) are greater than year (0.0%), market (3.77%) and CEO (16.67%) effects (p-values b 0.01). Alsonotable is the extent to which facility effects matter for non-profit hospitals, relative to other ownership types. Among for-profithospitals, facility effects (23.05%) also dominate year (7.59%), market (5.31%) and CEO (15.24%) effects (p-values b 0.01). Wenote that facility effects are greater for non-profits and for-profits than for public-direct and public-autonomous hospitals

7 Implemented using the “xtmixed” procedure in Stata 12.0.


(p-values b 0.01). While system effects are virtually non-existent among both types of public hospitals, they matter among thefor-profit hospitals and, to a lesser degree, the non-profit hospitals. Note, however, that the combined effect of bothorganization-level factors (facility and system) explains a substantial proportion of the variance among both for-profits (38.63%)and non-profits (46.21%). Prior studies have estimated that facility effects (i.e., business unit effects) explain between 29% and52% of the variance, and that system-level effects (i.e. corporate-parent effects) explain between 4% and 20% (Hough, 2006). Ourfindings with respect to private hospitals are consistent with these estimates and together further support hypothesis 2 and thegeneral dominance of organization-level factors among private hospitals.

While the evidence we present with respect to Hypotheses 1 and 2 stands on its own, it does not fully answer the keyquestion under consideration in this paper:When do leaders matter? Since the theorized logics and normative expectations arerelatively clear for public-direct, non-profit and for-profit hospitals, we cannot fully evaluate the role of these factors inexplaining the relative differences in class effects without comparison to circumstances where logics and normativeexpectations are more ambiguous. The results in Table 2 suggest a substantial CEO effect among public-autonomous hospitals.Importantly, this effect is statistically greater than the minimal effects estimated for year, market, system, and facility(p-values b 0.01). Thus, we find support for hypothesis 3a. Perhaps more importantly, the estimated CEO effect is statisticallygreater than the estimated effects for public-direct (p = 0.067), non-profit (p b 0.01) and for profit (p b 0.01) hospitals,supporting the idea that the ambiguous logic associated with the public-autonomous governance type may lend morediscretion and influence to the CEOs of these organizations. Thus, the results presented in Table 2 provide support forHypothesis 3b.

4.1. Robustness analysis

The findings we have presented may be sensitive to the choices we have made in developing our analysis. Specifically, ourfindings may be sensitive to the estimation method and performance measure. In order to address these concerns we ran anumber of robustness checks. First, while the results we have presented were produced using multi-level random effects,prior studies have relied on fixed effects techniques to partition the variance in firm performance (Mackey, 2008; McGahan &Porter, 2002). Accordingly, we examine the robustness of our findings to the use of a fixed effects model. Specifically, weestimated Eq. (1) using sequential ANOVA. While previous studies have employed slightly modified versions of ANOVA–nested ANOVA, simultaneous ANOVA–we employ sequential ANOVA for its simplicity and because prior studies have shownlittle difference across fixed effects methods (Hough, 2006; McGahan & Porter, 2002). All fixed effects ANOVA techniquesestimate variance components by sequentially running a series of “nested” models, beginning with a null model andsequentially adding each effect class from the highest level in the hierarchy to the lowest (in our case, we first enteredmarket, then system, then facility, then CEO and finally year), and noting the incremental change in R-squared at each level.The results of this analysis are presented in Table 3. For each governance type the table reports both the resulting sequentialsum of squares and the increment to R2 attributable to each effect class. As with our main findings, the results indicate thatthe largest CEO effect is observed among the public-autonomous hospitals (35.66%) and that this effect is significantlydifferent from the CEO effects estimated for public-direct (p b 0.05), non-profit (p b 0.01) and for-profit (p b 0.01) hospitals.Thus we find robust support for our key hypothesis, Hypothesis 3b. The results presented in Table 3 also supportHypotheses 1, 2 and 3a.

Second, while we have measured firm performance using ROA, we recognize that there are other ways to measure profitabilityand that performance might also be defined in other ways; for example productivity. Accordingly, we analyzed Eq. (1) whilereplacing ROA with two alternative measures of performance; one profit-based (total margin) and one productivity-based(adjusted length of stay). Total margin simply measures profitability as the ratio of net income–total revenues (operating andnon-operating) minus total expenses (operating and non-operating)–to total revenues (operating and non-operating). Adjustedlength of stay (ALOS) measures the average number of days a patient stays in the hospital, adjusted for the mix of services (some

Table 3Partitioning of variance in return on assets: fixed effects, sequential ANOVA.


n 381 217 1489 1046

Sequential sumof squares

Incrementto R2


IncrementR2


Incrementto R2


Incrementto R2

Year 1404.96 2.60% 9386.77 9.07% 1430.60 0.50% 6734.08 1.50%Market 4882.74 9.05% 38,325.53 37.03% 26,544.64 9.33% 26,022.83 5.78%System 2592.41 4.80% 846.80 0.82% 41,818.47 14.71% 74,476.85 16.55%Facility 6187.26 11.47% 7786.76 7.52% 89,627.87 31.52% 117,262.36 26.06%CEO 19,241.65 35.66% 25,384.94 24.53% 42,418.53 14.92% 94,286.92 20.95%Residual 21,055.61 39.02% 21,763.82 21.03% 83,963.78 29.53% 131,203.89 29.16%

Table 4Partitioning of variance in total margin: multi-level random effects.


n 379 215 1486 1049


Year b0.1 b0.01% b0.1 b0.01% b0.1 b0.01% b0.1 b0.01%Market 4.15 7.87% 118.36 47.38% 4.65 2.47% 29.20 10.60%System b0.1 b0.01% 6.24 2.50% 19.50 10.34% 49.69 18.04%Facility b0.1 b0.01% b0.1 b0.01% 85.56 45.37% 70.02 25.42%CEO 24.27 45.99% 50.43 20.19% 20.72 10.99% 45.62 16.56%Residual 24.35 46.14% 74.76 29.93% 58.13 30.83% 80.89 29.37%

⁎ Denotes the estimated % of total variance in total margin attributable to each effect class.


services are more intense than others) actually delivered by the hospital.8 Note that while the total margin model includes thesame hospital variables (debt-to-equity and patient volume) as the ROA model, our ALOS model replaces debt-to-equity with theshare of hospital patients on Medicare. This latter measure controls for the extent to which the hospital cares for elderly patientswho require more intense services (and thus might have longer ALOS).

The results of the model analyzing total margin are presented in Table 4. As with the results in Tables 2 and 3, the results inTable 4 indicate that the largest CEO effect is observed among the public-autonomous hospitals (45.99%) and that this effect issignificantly different from the CEO effects estimated for public-direct, non-profit, and for-profit (p-values b 0.01) hospitals. Thuswe find robust support for our key hypothesis, Hypothesis 3b. As with Table 3, we note that the results presented in Table 4 alsosupport Hypotheses 1, 2 and 3a.

The results of the model analyzing ALOS are presented in Table 5. Again we find that the largest CEO effect is observed amongthe public-autonomous hospitals (25.17%) and that this effect is significantly different from the CEO effects estimated forpublic-direct (p b 0.01), non-profit (p b 0.01) and for-profit (p b 0.01) hospitals. Thus we find robust support for our keyhypothesis, Hypothesis 3b. However, while the results reported support Hypothesis 2, there are some nuanced differencesreported in Table 5, specifically related to system and facility effects among public-direct and public-autonomous hospitals.Accordingly, the results in Table 5 do not support Hypothesis 1 or Hypothesis 3a. Such findings may be due to differences in howinstitutional logics affect efficiency relative to profitability or revenue considerations among public hospitals. Specifically, while apolitically oriented logic should be expected to center attention around the external determination of revenue flows andgovernment resource allocation processes, there is no conceptual reason to expect the same when it comes to the productivity ofoperational processes. Nevertheless, however the logic itself plays out, the results reported in Table 5 clearly suggest that CEOs ofpublic-autonomous hospitals appear to be less constrained than other types of hospitals in their influence over hospitalproductivity. Thus, along with the results reported in Tables 3 and 4, the results in Table 5 present clear and consistent evidencethat leaders matter more in circumstances characterized by ambiguous logics, where normative expectations are less clear.

5. Discussion

Our findings suggest interesting implications for the conversation surrounding whether or not–and if so, under whatconditions–CEOs “matter.” More specifically, they support the theoretical premise that ownership and governance arrangementscan serve as an important indirect contingency when it comes to CEO discretion. That is, beyond the direct restraining effect ofownership and governance, a clear institutional logic associated with certain ownership/governance arrangements can createnormative expectations for action that reduce leader discretion and, hence, mitigate leader effects. Consequently a “weak” (orambiguous) logic may be one circumstance under which CEOs matter more. In specific terms, our findings suggest that externaleffects are more prominent than organizational and CEO effects on financial performance for organizations whose ownership andgovernance arrangements are associated with an institutional logic that is externally-oriented, as in the case of hospitals withpublic-direct ownership/governance. Not surprisingly, year and market effects greatly exceeded those observed amongpublic-autonomous, non-profit and for-profit hospitals. Our results likewise show that, when the underlying institutional logic issimilarly clear yet narrowly centered on outcomes under the organization's control (rather than political and regulatoryexpediency)–as in private for-profit and non-profit hospitals–organizational effects explain more of the variance in profitabilitythan external and CEO effects. A key theoretical assumption underlying these results is that the locus of a particular institutionallogic (i.e., whether the “rules of the game” orient organizational members to things that are “external” to or “internal” to theorganization) is closely aligned with the relative influence of different levels of effects, and that regardless of locus, logics have acontingency effect on CEO discretion and thus shape the opportunity structure for a CEO's influence over performance.

Finally, and perhaps most notably, we found that when the dominant logic is, for all intents and purposes, “removed” from thepicture (i.e., when it becomes ambiguous and therefore “weak”—Mischel, 1977), CEO effects prevail. This particular finding wasconsistent for three different measures of performance, including two centered on profitability and one on productivity. Indeed,

8 This is done using a case mix index derived from Medicare's diagnostic related groups (DRGs).

Table 5Partitioning of variance in adjusted length of stay: multi-level random effects.


n 379 322 1719 1056


Year 0.82 11.86% 0.43 2.97% 0.70 6.99% b0.1 b0.01%Market b0.1 b0.01% 2.52 17.23% 1.19 11.89% 9.25 20.27%System 4.6635 67.20% b0.1 b0.01% b0.1 b0.01% 27.24 59.67%Facility 0.53 7.62% 9.30 63.50% 7.65 76.33% 5.27 11.56%CEO 1.75 25.17% 0.88 6.04% 0.40 4.02% 2.40 5.25%Residual b0.1 0.01% 1.50 10.26% 0.08 0.78% 1.48 3.25%

⁎ Denotes the estimated % of total variance in adjusted length of stay attributable to each effect class.


public-autonomous hospitals–whose ownership and governance structures comprise a mixture of public and private features–appear to allow CEOs the discretion needed to exert an effect on performance outcomes. That is, CEOs appear to be mostconsequential for performance when the organization operates under a structure that is associated with some measure of“logical” ambiguity (i.e., fuzziness regarding the appropriate field of action). CEOs in public-autonomous hospitals thus appear tohave more discretion and a stronger impact than in for-profit and non-profit hospitals, both of which have ownership andgovernance arrangements indicative of a clear, dominant institutional logic.

Not all our findings were expected. First, among public-direct hospitals, while CEOs did capture a smaller percentage ofvariance as compared to external effects, they were substantially higher than system and facility effects (25%, as indicated inTable 1). Large CEO effects may reflect the substantial role of response by top managers to political forces in the performance ofthese organizations. More specifically, it may be that performance among public-direct hospitals depends to a great degree on theability of a more extensive group of top and middle managers to navigate the political process and stem the influence ofpublic-choice on the behavior of those within the organization. In other words, there may be countervailing forces at work here,with externally situated logics and ownership demands limiting CEO discretion and the political capabilities of the larger group oftop and middle managers overcoming those restraints, to some extent. For example, when a frustrated county–the budgets ofwhich were being squeezed by various political realities–imposed a hard limit on its subsidies to the hospital it owned andgoverned, the CEO of the facility skillfully led a concerted effort to change the organization's culture and implement managed carecost-controls and quality improvements that resulted in improved revenue (Kane et al., 2012). That effort required the leader toskillfully navigate politically charged waters. As additional support for this idea, Carpenter, Sanders, and Gregersen (2001) foundthat U.S. multinational companies tend to perform better when the CEO has had international assignment experience, whichprovides the cultural and political “know-how” to navigate a foreign institutional environment.

The very small effect of the “system” level (i.e., the corporate-level effect) was likewise much smaller than we had expected itwould be (13.5% at its highest point, compared to up to 20% in prior studies). While results with respect to non-profit andfor-profit hospitals fit well within this range, the virtual absence of system effects among public hospitals of both types is striking.This finding may be partially explained by the extent to which public hospitals participate in multi-hospital systems. While nearly70% of private non-profit and for-profit hospitals in California have corporate parents, the numbers are smaller amongpublic-direct and public-autonomous hospitals; only 47% and 23%, respectively. Thus, among public hospitals, system (corporate)effects arise in less than half of all hospitals. Moreover, the theoretical explanations for why external influences dominate facilityand CEO effects among public-direct hospitals may apply equally to system effects. Specifically, when public choice and politicalforces are at play, the locus of control may be external to the organization in its entirety, including the system level.

Finally, although the degree to which CEOs “matter” appears to vary among different types of hospitals, still it is noteworthythat CEOs matter as much as they do, regardless of the organization's ownership and governance structures. This evidence hasintriguing implications for the debate on whether leaders matter, suggesting that CEOs have a hand in outcomes even in caseswhen such influence might be unexpected. Such leader effects may be even more compelling given that the CEO effects weidentify in our analysis are specifically the effect of CEO tenure (e.g., the years during which they were CEO), and not necessarilytheir overall impact on the organization. To the extent that CEOs are able to change enduring features of the organization(i.e., characteristics that persist beyond the leader's tenure) those effects would show up at the organization level (i.e., hospital orsystem). Thus, the results we report may underestimate CEO effects, to the extent that CEOs are able and willing to change theenduring features of their organizations. As has been noted by others (Blettner, Chaddad & Bettis, 2012), such limitations coupledwith our findings across different types of organizations bode well for future research on the mechanisms behind CEO effects inorganizations generally, and health care organizations specifically. In other words, the questions of exactly how leaders exert theeffect they do–central concerns of contingency theories in both leadership and strategic management literatures–appear tocontinue to be areas worthy of investigation, yet ones which were beyond the scope of this particular study.

Our research has additional implications for theory, research, and practice. First, our study fills an empirical gap in theliterature between micro (Klein et al., 2006) and macro (Crossland & Hambrick, 2007; Hambrick & Abrahamson, 1995) levelimpacts of leader discretion on organizational performance by providing evidence on the impact of organizational-level factors,namely, ownership and governance structure. Our findings thus extend contingency theories of leader effectiveness and strategicmanagement by identifying the clarity of institutional logics as determined by ownership and governance alignment as a key


locus upon which leadership effects may be contingent. Second, for organizational research, our findings suggest that controllingfor ownership type alone may be insufficient. Governance structure and its alignment with ownership structure (and the strengthof the implied corresponding institutional logic) is an important factor for which studies should account when examiningleadership and performance. Finally, in terms of practice, our research highlights the importance of recognizing ownership andgovernance arrangements and dominant logics when making critical decisions related to hiring, compensating, and training topmanagers. For example, intensive training and coaching may be most appropriate in organizations with weaker or ambiguouslogics and more managerial discretion; in organizations beholden to strong, politically-oriented institutional logics, however,more focused training that emphasizes engaging and navigating local political processes may be sufficient. The key idea here isthat leadership development processes should necessarily account for the nature of the organization's governance structure anddominant logic. In a similar vein, leaders themselves should be cognizant of the limits of their direct influence and of accountingfor the structure and logic of the organization. Such thinking might include tailoring the leadership approach (i.e., managementtools) to the normative expectations embedded in the organization.

Like all research, ours has limitations that point the way toward future studies. First and foremost, we recognize that this studydeals with CEO dynamics in only one industry (health care), drawing on data from one state. Future studies might confirm thesefindings among top management teams, in different regions, and in industries that exhibit similar variation in terms of ownershipand governance, and therefore dominant logics, among member organizations (e.g., universities, churches). Doing so may alsoallow consideration of additional performance measures, including quality metrics in industries for which financial performanceplays a relatively minor role. In addition, because our paper focuses on the question of whether and under what conditions CEOsmatter, our research does not inform the micro-level debate about which leader, follower, and/or task characteristics matter mostfor leader effectiveness. Specifically, by looking at changes in CEO without any information about CEO differences in personality,behavior, leader–follower relationship, or immediate task environment, our analysis does not allow us to draw conclusions aboutthe extent to which these factors matter. Thus, although we can conclude that the presence of different CEOs matters and thatcertain organizational-level conditions influence which CEOs matter more or less, we cannot draw conclusions about whethercertain leadership behaviors–i.e., the process of influence–matter or not and under what circumstances. Future research mightconsider both questions simultaneous to provide insight about the micro-aspects of leadership that matter most under varyingownership–governance arrangements and dominant logics.

6. Conclusion

The role of CEOs in organizational success (or failure) has been a longstanding interest of organizational scholars. BecauseCEOs may not be reasonably expected to be able to effectively shape their organizations in all cases and circumstances, we havesought to shed light on the conditions under which they appear to matter most for firm performance. The results of our researchsupport the idea that ownership and governance arrangements, and the ensuing strength of the situation, can be both a direct–power-based–and indirect–logic-based–contingent force affecting the influence of CEOs. Moreover, the focus of our theoreticalargument has been the idea that when logics are clear, they can be an impediment to the discretion of top managers, bindingthem to a limited range of actions and diluting managers' overall impact. In some cases, that logic may be oriented toward andcompounded by external forces (hence, the dominant influence of external effects among public-direct hospitals), while in others,the logic may correspond more directly to the outcomes pursued by the organization (hence, the dominant facility effects amongnon-profit and for-profit hospitals). Yet, even where ownership and governance serve as a constraining influence, it appears thatCEOs still exert a notable influence on the outcomes of their organizations. As such, much of the work of CEOs in such situationsmay lie in effectively navigating the constraints they face so as to redirect the organization in a desirable manner. To return to ouropening question, then, we might offer the tentative answer that ownership and governance structures indicative of anambiguous institutional logic represent one important circumstance under which CEOs are positioned to have greater influenceon the financial outcomes of their organizations, since such situations are theoretically associated with increased managerialdiscretion and hence provide a wider “window of opportunity” for leader effects.

Acknowledgments

We wish to thank Craig Crossland, Glen Kreiner, Vilmos Misangyi and the participants in the 2012 Academy of ManagementAnnual Meeting for their helpful and constructive feedback on the earlier versions of this paper.

References

Avolio, B. J., Waldman, D. A., & Einstein, W. O. (1988). Transformational leadership in a management game simulation. Group and Organizational Studies, 13, 59–80.Barrick, M. R., & Mount, M. K. (1993). Autonomy as a moderator of the relationships between the Big Five personality dimensions and job performance. Journal of

Applied Psychology, 78(1), 111–118.Bertrand, M., & Schoar, A. (2003). Managing with style: The effect of managers on firm policies. Quarterly Journal of Economics, 118, 1169–1208.Blettner, D. P., Chaddad, F. R., & Bettis, R. A. (2012). The CEO performance effect: Statistical issues and a complex fit perspective. Strategic Management Journal,

33(8), 986–999.Bobko, P. (2001). Correlation and regression: Applications for industrial organizational psychology and management. Thousand Oaks, CA: Sage.Buchanan, J. M. (1972). Theory of public choice. Ann Arbor: University of Michigan Press.Carpenter, M. A., Sanders, W. G., & Gregersen, H. B. (2001). Bundling human capital with organizational context: The impact of international assignment

experience on multinational firm performance and CEO pay. Academy of Management Journal, 493–511.

http://refhub.elsevier.com/S1048-9843(13)00101-X/rf0010











Colbert, A. E., Kristof-Brown, A. L., Bradley, B. H., & Barrick, M. R. (2008). CEO transformational leadership: The role of goal importance congruence in topmanagement teams. Academy of Management Journal, 51, 81–96.

Corley, K. G., & Gioia, D. A. (2004). Identity ambiguity and change in the wake of a corporate spin-off. Administrative Science Quarterly, 49(2), 173–208.Crossland, C., & Hambrick, D. (2007). How national systems differ in their constraints on corporate executives: A study of CEO effects in three countries. Strategic

Management Journal, 28, 767–789.Crossland, C., & Hambrick, D. (2011). Differences in managerial discretion across countries: How nation-level institutions affect the degree to which CEOs matter.

Strategic Management Journal, 32(8).Cuervo, A., & Villalonga, B. (2000). Explaining the variance in the performance effects of privatization. Academy of Management Review, 25, 581–590.Day, D. V., & Lord, R. G. (1988). Executive leadership and organizational performance: Suggestions for a new theory and methodology. Journal of Management, 14,

453–464.DiMaggio, P. J., & Powell, W. W. (1983). The iron cage revisited: Institutional isomorphism and collective rationality in organizational fields. American Sociological

Review, 48, 147–160.Dionne, S. D., Yammarino, F. J., Howell, J. P., & Villa, J. (2005). Substitutes for leadership, or not. The Leadership Quarterly, 16(1), 169–193.Dutton, J. E., & Dukerich, J. M. (1991). Keeping an eye on the mirror: Image and identity in organizational adaptation. Academy of Management Journal, 34,

517–554.Dutton, J. E., & Penner, W. J. (1993). The importance of organizational identity for strategic agenda building. In J. Hendry, & G. Johnson (Eds.), Strategic thinking:

Leadership in the management of change (pp. 89–113). New York: Strategic Management Society, Wiley.Dvir, T., & Shamir, B. (2003). Follower developmental characteristics as predicting transformational leadership: A longitudinal field study. The Leadership

Quarterly, 14(3), 327–344.Eccles, R. G., & Nohria, N. (1992). Beyond the hype: Rediscovering the essence of management. Boston: Harvard Business School Press.Fiedler, F. E. (1967). A theory of leadership effectiveness. New York: McGraw-Hill.Finkelstein, S., & Hambrick, D. C. (1990). Top-management-team tenure and organizational out-comes: The moderating role of managerial discretion.

Administrative Science Quarterly, 35, 484–503.Friedland, R., & Alford, R. R. (1991). Bridging society back in: Symbols, practices, and institutional contradictions. In W. W. Powell, & P. J. DiMaggio (Eds.), The new

institutionalism in organizational analysis (pp. 232–263). Chicago: University of Chicago Press.Glaeser, E. L. (2003). Introduction. In E. Glaeser (Ed.), The governance of not-for-profit organizations (pp. 1–44). Chicago, IL: University of Chicago Press.Goodstein, J., & Boeker, W. (1991). Turbulence at the top: A new perspective on governance structure changes and strategic change. Academy of Management

Journal, 34(2), 306–330.Hambrick, D. C., & Abrahamson, E. (1995). Assessing managerial discretion across industries: A multimethod approach. Academy of Management Journal, 38(5),

1427–1441.Hambrick, D., & Finkelstein, S. (1987). Managerial discretion: A bridge between polar views of organizational outcomes. In B. Staw, & L. Cummings (Eds.), Research

in organizational behavior, vol. 9. (pp. 369–406)PressGreenwich, CT: JAI.Hambrick, D. C., & Mason, P. A. (1984). Upper echelons: The organization as a reflection of its top managers. Academy of Management Review, 9(2), 193–206.Hannan, M., & Freeman, J. (1977). The population ecology of organizations. American Journal of Sociology, 82, 929–964.Herrmann, D., & Felfe, J. (2013). Moderators of the relationship between leadership style and employee creativity: The role of task novelty and personal initiative.

Creativity Research Journal, 25(2), 172–181.Hough, J. R. (2006). Business segment performance redux: A multi-level approach. Strategic Management Journal, 27, 45–61.House, R. (1971). A path goal theory of leader effectiveness. Administrative Science Quarterly, 16(3), 321–339.House, R. (1996). Path–goal theory of leadership: Lessons, legacy and a reformulated theory. Leadership Quarterly, 7(3).House, R. J., Spangler, W. D., & Woycke, J. (1991). Personality and charisma in the U.S. presidency: A psychological theory of leader effectiveness. Administrative

Science Quarterly, 36, 364–396.Hutzschenreuter, T., Kleindienst, I., & Greger, C. (2012). How new leaders affect strategic change following a succession event: A critical review of the literature.

Leadership Quarterly, 23(5), 729–755.Jackall, R. (1988). Moral mazes: The world of corporate managers. New York: Oxford University Press.Jones, B. F., & Olken, B. A. (2005). Do leaders matter? National leadership and growth since World War II. The Quarterly Journal of Economics, 120(3), 835–864.Kane, N., Singer, S., Clark, J., Eckloo, K., & Valentine, M. (2012). Strained local and state government finances among current realities that threaten public hospitals'

profitability. Health Affairs, 31(8), 1680–1689.Kellerman, B. (2008). Followership: How followers are creating change and changing leaders. Boston, MA: Harvard Business School Press.Kerr, S., & Jermier, J. M. (1978). Substitutes for leadership: Their meaning and measurement. Organizational Behavior and Human Performance, 22, 375–403.Klein, K. J., Ziegert, J. C., Knight, A. P., & Xiao, Y. (2006). Dynamic delegation: Shared, hierarchical, and deindividualized leadership in extreme action teams.

Administrative Science Quarterly, 51(4), 590–621.Lieberson, S., & O'Connor, J. (1972). Leadership and organizational performance: A study of large corporations. American Sociological Review, 37, 117–130.Mackey, A. (2008). The effect of CEOs on firm performance. Strategic Management Journal, 29, 1357–1367.McAdam, Doug (1996). Conceptual origins, current problems, future directions. In Doug McAdam, John D. McCarthy, & Mayer N. Zald (Eds.), Comparative

perspectives on social movements (pp. 23–40). Cambridge: Cambridge University Press.McGahan, A. M., & Porter, M. E. (2002). What do we know about variance in accounting profitability. Management Science, 48(7), 834–851.Misangyi, V., Elms, H., Greckhamer, T., & Lepine, J. (2006). A new perspective on a fundamental debate: A multilevel approach to industry, corporate, and business

unit effects. Strategic Management Journal, 27, 571–590.Mischel, W. (1977). The interaction of person and situation. In D. Magnusson, & N. S. Endler (Eds.), Personality at the crossroads: Current issues in interactional

psychology (pp. 333–352). Hillsdale, NJ: Erlbaum.Moe, T. M. (1990). The politics of structural choice: Toward a theory of public bureaucracy. In O. E. Williamson (Ed.), Organization theory: From Chester Barnard to

the present and beyond (pp. 116–153). New York, NY: Oxford University Press.Nübold, Annika, Muck, Peter M., & Maier, Günter W. (2012). A new substitute for leadership? Followers' state core self-evaluations. The Leadership Quarterly,

24(1), 29–44.Ocasio, W. (1997). Towards an attention-based view of the firm. Strategic Management Journal, 18, 187–206 (Summer Special Issue).Peteraf, M., & Reed, R. (2007). Managerial discretion and internal alignment under regulatory constraints and change. Strategic Management Journal, 28(11),

1089–1112.Purvanova, R. K., & Bono, J. E. (2009). Transformational leadership in context: Face-to-face and virtual teams. Leadership Quarterly, 20, 343–357.Salancik, G., & Pfeffer, J. (1977). Constraints on administrator discretion: The limited influence of mayors on city budgets. Urban Affairs Quarterly, 12, 475–498.Shamir, B., & Howell, J. M. (1999). Organizational and contextual influences on the emergence and effectiveness of charismatic leadership. The Leadership

Quarterly, 10(2), 257–283.Shamir, B., & Howell, J. M. (2000). The role of followers in the charismatic leadership process: Susceptibility, social construction, and leader empowerment. Paper

presented at the Academy of Management Annual Meeting, Toronto.Sutherland, J., Fisher, E., & Skinner, J. (2009). Getting past denial—The high cost of health care in the united states. New England Journal of Medicine, 361, 1227–1230.Thomas, A. (1988). Does leadership make a difference to organizational performance? Administrative Science Quarterly, 33, 388–400.Thornton, P. (2002). The rise of the corporation in a craft industry: Conflict and conformity in institutional logics. Academy of Management Journal, 45, 81–101.Thornton, P. H., & Ocasio, W. (1999). Institutional logics and the historical contingency of power in organizations: Executive succession in the higher education

publishing industry, 1958–1990. American Journal of Sociology, 105(3).Thornton, P. H., & Ocasio, W. (2008). Institutional logics. In R. Greenwood, C. Oliver, K. Sahlin, & R. Suddaby (Eds.), The Sage handbook of organizational

institutionalism. Thousand Oaks: Sage Publications.


















































































Tsui, A. S., Zhang, Z. X., Wang, H., Xin, K. R., & Wu, J. B. (2006). Unpacking the relationship between CEO leadership behavior and organizational culture. TheLeadership Quarterly, 17, 113–137.

Vitaliano, D. (2003). Do not-for-profit firms maximize profit? The Quarterly Review of Economics and Finance, 43(1), 75–87.Voss, Z. G., Cable, D. M., & Voss, G. B. (2006). Organizational identity and firm performance: What happens when leaders disagree about “who we are”?

Organization Science, 17, 741–755.Waldman, D. A., Ramirez, G. A., House, R. J., & Puranam, P. (2001). Does leadership matter? CEO leadership attributes and profitability under conditions of

perceived environmental uncertainty. Academy of Management Journal, 44, 134–143.Weiner, N. (1978). Situational and leadership influences on organizational performance. Proceedings of the Academy of Management, 230–234.Wennberg, J., & Gittelsohn, A. (1973). Small area variations in health care delivery. Science, 182, 1102–1108.Whetten, D. A., & Godfrey, P. C. (1998). Identity in organizations: Building theory through conversations. Thousand Oaks, CA: Sage.











when do leaders matter? ownership, governance and the influence of ceos on firm performance

Documents