e psycal

Matthew Hall *

ity and psychological empowerment. This result highlights the role of cognitive and motivational mechanisms inexplaining the eect of management accounting systems on managerial performance. In particular, the results indicate

hensive PMS include a more diverse set of perfor-mance measures, and performance measures thatare linked to the strategy of the rm and provideinformation about parts of the value chain (Chen-hall, 2005; Malina & Selto, 2001; Nanni, Dixon, &

eserved.

* Tel.: +44 0 20 7955 7736; fax: +44 0 20 7955 7420.E-mail address: M.R.Hall@lse.ac.uk

Available online at www.sciencedirect.com

Accounting, Organizations and Soci0361-3682/$ - see front matter 2007 Elsevier Ltd. All rights rthat comprehensive PMS inuences managers cognition and motivation, which, in turn, inuence managerialperformance. 2007 Elsevier Ltd. All rights reserved.

Introduction

In recent years organizations have sought todevelop more comprehensive performance mea-surement systems (PMS) to provide managers

and employees with information to assist in man-aging their rms operations (Fullerton & McWat-ters, 2002; Ittner, Larcker, & Randall, 2003; Lillis,2002; Malina & Selto, 2001; Ullrich & Tuttle,2004). Prior research indicates that more compre-Department of Accounting and Finance, London School of Economics and Political Science, Houghton Street,

London WC2A 2AE, United Kingdom

Abstract

This study examines how comprehensive performance measurement systems (PMS) aect managerial performance.It is proposed that the eect of comprehensive PMS on managerial performance is indirect through the mediating vari-ables of role clarity and psychological empowerment. Data collected from a survey of 83 strategic business unit man-agers are used to test the model. Results from a structural model tested using Partial Least Squares regression indicatethat comprehensive PMS is indirectly related to managerial performance through the intervening variables of role clar-The eect of comprehensivsystems on role clarity, p

and manageridoi:10.1016/j.aos.2007.02.004erformance measurementhological empowermentperformance

www.elsevier.com/locate/aos

ety 33 (2008) 141163

tionsVollman, 1992; Neely, Gregory, & Platts, 1995).Comprehensive PMS have been popularised intechniques such as the balanced scorecard (Kaplan& Norton, 1996), tableau de bord (Epstein &Manzoni, 1998) and performance hierarchies(Lynch & Cross, 1992).

In this paper I examine how comprehensive PMSaect managerial performance. Prior research hasfocused on the relation between comprehensivePMS and organisational performance (perceivedor actual) (Chenhall, 2005; Davis & Albright,2004; Hoque & James, 2000; Ittner, Larcker, &Randall, 2003; Said, HassabElnaby, & Wier,2003), and on the use of multiple performance mea-sures in performance evaluation judgements(Banker, Chang, & Pizzini, 2004; Lipe & Salterio,2000; Schi & Homan, 1996). However, there islimited empirical research that examines the behav-ioural consequences of comprehensive PMS (Ittner& Larcker, 1998; Webb, 2004). Studies examininglinks between management control systems andorganisational outcomes assume that such systemsaect the behaviour of individuals within the orga-nization, which then facilitates the achievement oforganisational goals. However, as Chenhall (2003)notes, this assumption involves broad leaps in logicand there is no compelling evidence to suggest thatthese links exist. Similarly, Covaleski, Evans, Luft,and Shields (2003) argue that studies at the organi-sational level of analysis remain somewhat limitedbecause they are based upon assumptions about,rather than a detailed investigation of, individualbehaviour.

Further, there is little empirical research thatexamines whether control system componentshave direct and/or indirect eects on work perfor-mance (Shields, Deng, & Kato, 2000). This isimportant because there can be theoretical dier-ences between direct- and indirect-eects modelsthat can have practical implications (Shieldset al., 2000). Psychological theories indicate thatcognitive and motivational mechanisms are likelyto explain the relation between comprehensivePMS and managerial performance (Collins, 1982;Ilgen, Fisher, & Taylor, 1979; Luckett & Eggleton,1991). As such, I examine how the relationbetween comprehensive PMS and managerial

142 M. Hall / Accounting, Organizaperformance can be explained by the interveningvariables of role clarity and psychologicalempowerment.

Recent research indicates that the informationdimensions of management accounting practices,such as PMS, are not captured eectively by labelssuch as the balanced scorecard (Chenhall, 2005;Ittner, Larcker, & Randall, 2003). In particular,Ittner, Larcker, and Randall (2003) argue thatresearchers need to devise improved methods fordetermining what rms mean by contemporaryPMS. As such, in this study, I draw on descrip-tions of PMS from the performance measurementliterature to develop a denition of a comprehen-sive PMS. Based on this denition, I develop aninstrument to measure empirically the comprehen-sive PMS construct.

Data collected from a survey of strategic busi-ness unit (SBU) managers are used to examinehow comprehensive PMS is related to managerialperformance. I focus on SBU managers as theinformation provided by comprehensive PMS isexpected to be useful at this managerial levelbecause of SBU managers information require-ments. The results show that comprehensivePMS is indirectly related to managerial perfor-mance through the intervening variables of roleclarity and psychological empowerment. Consis-tent with theory, the results highlight the role ofcognitive and motivational mechanisms inexplaining the eect of management accountingsystems on managerial performance. In particular,the results indicate that comprehensive PMS inu-ences managers cognition and motivation, which,in turn, inuence managerial performance. Thiscontributes to prior research that has examinedthe direct and indirect eects of management con-trol systems on work performance (Shields et al.,2000), and also extends the limited body of priorresearch that has examined the eect of manage-ment control system attributes on psychologicalempowerment (Smith & Langeld-Smith, 2003;Spreitzer, 1995, 1996) and role clarity (Chenhall &Brownell, 1988). Finally, the study responds tocalls to develop improved methods for examin-ing the attributes of management accountingpractices by developing a reliable and valid ins-trument to measure the comprehensive PMS

and Society 33 (2008) 141163construct.

integration of measures with strategy and across

empowerment (H5).

Comprehensive performance measurement systems

Recent research has emphasised the importanceof examining the information dimensions of con-temporary PMS (Chenhall, 2005; Ittner, Larcker, &Randall, 2003; Luft & Shields, 2003). The per-formance measurement literature has identiedseveral important characteristics of more compre-hensive PMS. Malina and Selto (2001) argue thata comprehensive PMS consist of a parsimoniousset of critical performance measures. Results oftheir study show that the balanced scorecard wasconsidered comprehensive when it provided anoverall measure of business performance. Onemanager stated that the BSC is trying to give usa broader business set of measures of success thanmore traditional nancial or market share. Itwraps a set of things together that makes sense

tionsthe value chain) is expected to be useful for SBUmanagers because their jobs require considerationof multiple aspects of the SBUs operations andconsideration of strategic issues. Thus, compre-hensive PMS is expected to provide importantinformation for SBU managers to enhance theirrole clarity and psychological empowerment,and, in turn, enhance managerial performance.

The theoretical model is shown in Fig. 1. For theThe remainder of the paper contains four sec-tions: the next section develops the theoreticalmodel, including presentation of the hypotheses.The research method, including sample selectionand variable measurement, is then presented. Thisis followed by presentation of the results. The nalsection discusses the results and concludes thepaper.

Theoretical development and hypotheses

formulation

A major premise behind the development ofmore comprehensive PMS is that they can helpto improve managerial performance (Atkinson &Epstein, 2000; Epstein & Manzoni, 1998; Kaplan &Norton, 1996). Psychological theories indicatethat cognitive and motivational mechanisms arelikely to explain the relation between comprehen-sive PMS and managerial performance (Ilgenet al., 1979). As such, comprehensive PMS is notexpected to have a direct eect on managerial per-formance. Rather, comprehensive PMS is expectedto have an indirect eect on managerial perfor-mance by: (1) clarifying managers role expecta-tions, and (2) providing feedback to enhancemanagers intrinsic task motivation (Collins,1982; Luckett & Eggleton, 1991). Thus, theorypredicts that role clarity and psychologicalempowerment are likely to mediate the relationbetween comprehensive PMS and managerial per-formance. In particular, comprehensive PMS areexpected to have positive eects on SBU managersbehaviour. This is because the information pro-vided by comprehensive PMS (information aboutthe important parts of the SBUs operations, and

M. Hall / Accounting, Organizarole clarity path, I argue that comprehensive PMSenhances role clarity (H1), and role clarity enhancesmanagerial performance (H2). For the psychologi-cal empowerment path, I argue that comprehensivePMS enhances psychological empowerment (H3),and psychological empowerment enhances mana-gerial performance (H4). I also propose a positiveassociation between role clarity and psychological

H2

H5

H1

H4

H3

Managerial Performance

Role clarity

ComprehensivePMS

Psychologicalempowerment

Fig. 1. Theoretical model: comprehensive PMS, role clarity,psychological empowerment and managerial performance.

and Society 33 (2008) 141163 143for managing the business (Malina & Selto,

such, a more comprehensive PMS is one that pro-vides more comprehensive performance informa-tion to managers, i.e., measures that fullydescribe the SBUs operations and link to strategyand across the value chain. In contrast, a less com-prehensive PMS is one that provides less compre-

tions and Society 33 (2008) 1411632001, p. 70). Ittner, Larcker, and Randall (2003)argue that a broad set of measures, or measure-ment diversity, is an important feature of morecomprehensive PMS. Ittner, Larcker, and Randall(2003, p. 717) consider measurement diversity assupplementing traditional nancial measureswith a diverse mix of non-nancial measures thatare expected to capture key strategic performancedimensions that are not accurately reected inshort-term accounting measures. Similarly, Ull-rich and Tuttle (2004) and Henri (2006) argue thatcomprehensive systems are designed to measureperformance in all the important areas of the rm.These studies indicate that providing a broad set ofmeasures that cover dierent parts of the organiza-tions operations is an important aspect of morecomprehensive PMS.

The integration of measures with strategy andproviding information about parts of the valuechain is also an important feature of more compre-hensive PMS. Nanni et al. (1992) argue that PMSthat integrate actions across functional boundaries,and focus on strategic results, are critical in sup-porting the new manufacturing and competitiveenvironments facing organizations. In addition,the integration of measures with the strategy andobjectives of the organization provides perfor-mance information about progress on importantdimensions of performance (Kaplan & Norton,1996; Malina & Selto, 2001; Malmi, 2001; Nanniet al., 1992; Neely et al., 1995; Webb, 2004). Morecomprehensive PMS provide an understanding ofthe linkages between business operations and strat-egy (Chenhall, 2005).

Thus, the PMS literature indicates that there areseveral important characteristics of comprehensivePMS, including providing a broad set of measuresrelated to the important parts of the organisation,the integration of measures with strategy and val-ued organisational outcomes, and the integrationof measures across functional boundaries and thevalue chain (Chenhall, 2005; Henri, 2006; Ittner,Larcker, & Randall, 2003; Malina & Selto, 2001;Malmi, 2001; Neely et al., 1995). Therefore, it isargued that a comprehensive PMS provides per-formance measures that describe the importantparts of the SBUs operations and integrates mea-

144 M. Hall / Accounting, Organizasures with strategy and across the value chain. Ashensive performance information to managers,i.e., measures that only partially describe theSBUs operations and contain few (if any) linksto strategy and across the value chain.

The way in which comprehensive PMS provideenhanced performance information supplies thebasis for linking comprehensive PMS with SBUmanagers role clarity and psychological empower-ment. Individuals at higher levels in the organisa-tion, such as SBU managers, obtain feedbackabout the results of operations and work-relatedperformance from PMS (Collins, 1982; Luckett& Eggleton, 1991). A more comprehensive PMSprovides richer and more complete feedback aboutoperations and results to SBU managers (Chen-hall, 2005; Kaplan & Norton, 2001; Malina &Selto, 2001), which is expected to have positiveeects on managers role clarity and psychologicalempowerment.

Comprehensive PMS and role clarity

Role clarity refers to individuals beliefs aboutthe expectations and behaviours associated withtheir work role (Kahn, Wolfe, Quinn, Snoek, &Rosenthal, 1964).1 In this study I examine whethercomprehensive PMS is related to two aspects ofrole clarity; goal clarity (the extent to which theoutcome goals and objectives of the job are clearlystated and well dened) and process clarity (theextent to which the individual is certain abouthow to perform his or her job) (Sawyer, 1992). Itis expected that more comprehensive performanceinformation will help to clarify SBU managers

1 Kahn et al. (1964) use the term role ambiguity, which refersto uncertainty regarding parts of an individuals role. In thisstudy the term role clarity is used. However, this is conceptuallyno dierent from role ambiguity (Sawyer, 1992). Role clarity isexpressed as the extent of certainty, rather than ambiguity, of

role expectations.

tionsrole expectations and the appropriate behavioursfor fullling those role expectations.

Several researchers argue that more comprehen-sive performance information can help to improverole clarity. Collins (1982) argues that managementaccounting systems can be used to inform individ-uals about what is expected of them in their role.Specically, comprehensive performance informa-tion can serve to clarify individuals roles in theorganisation by making specic the goals andappropriate behaviours associated with a work role(Ilgen et al., 1979).

Comprehensive PMS can increase SBU manag-ers goal clarity by providing information aboutthe organizations strategies and operations, whichhelps them to better understand their own rolewithin the organization. Access to comprehensiveperformance information allows SBU managersto see the big picture and develop a referencepoint for understanding their roles within theirorganization (Bowen & Lawler, 1992; Lawler,1992). More comprehensive PMS can help to clar-ify and communicate strategic intent, and can cap-ture dierent dimensions of performance, which isimportant in describing the organizations opera-tions (Kaplan & Norton, 1996; Lynch & Cross,1992; Simons, 2000). Performance feedback aboutbusiness unit operations increases managers levelof certainty over the requirements of their workrole (Kahn et al., 1964; King & King, 1990). Assuch, more comprehensive PMS should improveSBU managers understanding of what comprisestheir role and what is expected of them, and thusincrease goal clarity.

Comprehensive PMS can increase process clar-ity by providing performance information toimprove SBU managers understanding of thedrivers of performance, the eect of their actionson parts of the value chain, and the links betweendierent parts of the organizations operations. Inparticular, more comprehensive PMS can educateSBU managers about the economics of the busi-ness and the drivers of costs, revenues and perfor-mance (Kaplan & Norton, 1996; Lynch & Cross,1992; Simons, 2000). Banker et al. (2004) arguethat the integration of measures across the valuechain can help individuals to understand cross-

M. Hall / Accounting, Organizafunctional relationships. Similarly, Malina andSelto (2001) found that the balanced scorecardwas important for managing the business whenperformance information was comprehensive andintegrated. As such, more comprehensiveperformance information is expected to improveSBU managers understanding of their workrole and thus increase role clarity, which leadsto H1.

H1: There is a positive relation between com-prehensive PMS and role clarity.

Role clarity and managerial performance

Individuals require sucient information toperform tasks eectively. A lack of informationregarding the goals of the job and the most eec-tive job behaviours can result in eort that is inef-cient, misdirected or insucient for the task(s),and thus reduce job performance (Jackson &Schuler, 1985; Tubre & Collins, 2000). SBU man-agers are likely to be more eective when theyunderstand what needs to be done and how man-agerial functions are to be performed. Empiricalresults indicate that role ambiguity decreases workperformance (Abramis, 1994; Jackson & Schuler,1985; Tubre & Collins, 2000). These argumentsand evidence lead to H2:

H2: There is a positive relation between roleclarity and managerial performance.

Comprehensive PMS and psychological

empowerment

Psychological empowerment refers to increasedintrinsic task motivation manifested in a set offour cognitions; meaning (the value placed onwork judged in relation to an individuals own ide-als or standards), competence (an individualsbelief in his/her capacity to perform a job withskill), self-determination (an individuals beliefconcerning the degree of choice they have in initi-ating and performing work behaviours), andimpact (the extent to which an individual believes

and Society 33 (2008) 141163 145they can inuence outcomes at work) (Spreitzer,

1995; Thomas & Velthouse, 1990).2 Higher levelsof meaning, competence, self-determination andimpact reect higher intrinsic task motivation(Thomas & Velthouse, 1990), and, therefore, areexpected to result in more focused attention ontasks, greater eort (intensity) and persistence dur-

146 M. Hall / Accounting, Organizationsing tasks, and improved task strategies (Mitchell &Daniels, 2003; Pinder, 1998).

Providing adequate performance informationenhances the development of psychologicalempowerment. Feedback theories from psychol-ogy indicate that performance information canimprove psychological empowerment by providinginformation about task behaviour and perfor-mance (Collins, 1982; Ilgen et al., 1979; Locke,Shaw, Saari, & Latham, 1981; Luckett & Eggleton,1991). In particular, intrinsic task motivation isincreased when managers are provided with feed-back about the results of operations (Ilgen et al.,1979). The greater the amount of information pro-vided on a job, the greater will be the motivatingpotential of the job (Ilgen et al., 1979). This isbecause performing a task without knowledge ofresults provides little feedback to managers, whichis likely to be frustrating and dissatisfying, thusreducing intrinsic motivation (Luckett & Eggleton,1991).

The empowerment literature also supports thelink between performance information and intrin-sic motivation. Providing information about theperformance of the business is essential for thedevelopment of empowerment (Bowen & Lawler,1992; Spreitzer, 1995, 1996; Quinn & Spreitzer,1997). In contrast, a lack of information aboutperformance has adverse aects on feelings ofempowerment (Chiles & Zorn, 1995). In supportof these arguments, Spreitzer (1995, 1996) foundthat access to cost and quality performance infor-mation is positively associated with psychologicalempowerment.

2 Psychological empowerment is a motivational construct andis therefore distinguished from objective structural factors, suchas delegation of decision-making authority (Thomas & Velt-house, 1990). Delegation is likely to enhance psychologicalempowerment; however, it is individuals cognitive interpreta-tions of such structural factors that leads to stronger psycho-logical empowerment, rather than some objective reality (Chiles

& Zorn, 1995; Spreitzer, 1996; Thomas & Velthouse, 1990).As such, SBU managers require informationabout the results of SBU operations to feel intrin-sically motivated. The characteristics of compre-hensive PMS (providing performance measuresthat describe the important parts of the SBUsoperations and integrating measures with strategyand across the value chain) provide a rich and rel-atively complete picture of the performance of thebusiness unit (Chenhall, 2005; Ittner, Larcker, &Randall, 2003; Kaplan & Norton, 2001; Malina& Selto, 2001). Such information is essentialfor SBU managers because their jobs requireconsideration of multiple aspects of the SBUsoperations and consideration of strategic issues.As such, a more comprehensive PMS providesthe performance information necessary for SBUmanagers to develop higher levels of psychologicalempowerment. In contrast, a less comprehensivePMS provides limited and inadequate perfor-mance information, and thus is likely to limit thedevelopment of SBU managers psychologicalempowerment.

Comprehensive PMS is expected to increaseSBU managers beliefs regarding each dimensionof psychological empowerment: meaning, compe-tence, self-determination and impact. Congerand Kanungo (1988) argue that performanceinformation is likely to strengthen individualsbeliefs of meaning and purpose, as managersbelieve they are valued when they are providedwith the results of operations. Further, Spreitzer(1995) argues that greater access to performanceinformation is essential in enabling managers tobelieve that their work is valuable. A more com-prehensive PMS provides a rich and relativelycomplete picture of the performance of the busi-ness units operations, which increases SBU man-agers ability to judge the value of their work inthe context of the organizations strategies andoperations. As such, a more comprehensive PMScan make SBU managers believe their jobs aremore meaningful by helping them to determinehow their work ts within the broader scope ofthe organization. Without comprehensive infor-mation about performance, SBU managers arelikely to place little value on their work withinthe organization, and thus experience lower levels

and Society 33 (2008) 141163of meaning.

tionsGist and Mitchell (1992) argue that competencebeliefs are strengthened by providing performanceinformation to individuals in the organisation.This is because performance information improvesindividuals ability to make assessments of theirperformance capabilities. By providing informa-tion about business unit operations, and links tostrategy and the value chain, a more comprehen-sive PMS provides improved knowledge of results,which is fundamental for reinforcing a sense ofcompetence (Gist & Mitchell, 1992; Ilgen et al.,1979; Lawler, 1992; Spreitzer, 1995). A less com-prehensive PMS provides inadequate knowledgeof results, and therefore reduces SBU managersbelief in their ability to perform tasks competently(Conger & Kanungo, 1988; Thomas & Velthouse,1990).

Comprehensive PMS is expected to increaseself-determination. SBU managers require infor-mation about where their organization is headedin order to believe they are capable of taking theinitiative (Kanter, 1989). Adequate knowledge ofresults is essential for managers to be able to directand manage their own performance (Lawler,1992). Managers need to understand how welltheir business unit is performing to be condentenough to make decisions on their own (Spreitzer,1995). A more comprehensive PMS provides a richand relatively complete picture of the businessunits performance, which increases SBU manag-ers condence to initiate and complete tasks ontheir own, thus increasing self-determination. Aless comprehensive PMS provides inadequate per-formance information, and thus reduces SBUmanagers condence to initiate and regulate theirown actions.

Comprehensive PMS is also expected toincrease impact. To have an impact, managersneed to understand how their business unit is per-forming (Spreitzer, 1995). Further, managersrequire adequate performance information inorder to believe they can make and inuence deci-sions that are consistent with the organizationspriorities (Lawler, 1992). A more comprehensivePMS strengthens SBU managers knowledge ofoperations and organisational priorities, andtherefore improves managers ability to inuence

M. Hall / Accounting, Organizaand act in ways that are consistent with those pri-orities, thus increasing impact. In contrast, a lesscomprehensive PMS provides limited knowledgeof organisational priorities and strategies. Withoutsucient knowledge of results, managers are unli-kely to exert inuence in their work area (Kraimer,Seibert, & Liden, 1999).

In summary, comprehensive PMS is expected tobe positively related to each dimension of psycho-logical empowerment, which leads to H3:

H3: There is a positive relation between com-prehensive PMS and the four dimensions ofpsychological empowerment.

Psychological empowerment and managerial

performance

Empowered individuals should perform betterthan those individuals who are less empowered(Liden, Wayne, & Sparrowe, 2000). This is becauseempowerment increases both initiation and persis-tence of managers task behaviour (Conger &Kanungo, 1988; Thomas & Velthouse, 1990). Inparticular, higher levels of psychological empower-ment lead to greater eort and intensity of eort,persistence, and exibility (Spreitzer, 1995; Tho-mas & Velthouse, 1990), all of which are behav-iours that enhance performance (Mitchell &Daniels, 2003; Pinder, 1998).

Each dimension of psychological empowermentis related to behaviours that enhance managerialperformance. Individuals who place more mean-ing, or care more, about their work put forth moreeort and are more committed to their tasks, andthus likely to persist in the face of obstacles or set-backs (Kanter, 1983; Liden et al., 2000; Thomas &Velthouse, 1990). Individuals who believe they canperform well on a task (i.e., feel competent) do bet-ter than those individuals who think they will fail(Gist & Mitchell, 1992). Competence results inmore eort, persistence in the face of obstacles,and more initiative (Bandura, 1977; Spreitzer,Kizilos, & Nason, 1997; Thomas & Velthouse,1990). Spreitzer et al. (1997) and Liden et al.(2000) found that competence was positively asso-ciated with work performance. Self-determination

and Society 33 (2008) 141163 147results in more eort and persistence, and greater

managers to determine and take actions to com-plete tasks, and thus should increase self-determi-nation. A lack of role clarity is likely to makeindividuals believe they are helpless and thusreduce the impact they have in their work area

tered to SBU managers within Australian manu-facturing organizations. I obtained a list of 1000

tions and Society 33 (2008) 141163exibility to adapt to changing situations and cre-ate improved task strategies (Deci & Ryan, 1987;Thomas & Velthouse, 1990). Work performanceis enhanced when managers believe they haveautonomy over how their work is to be accom-plished (Miller & Monge, 1986). In relation toimpact, individuals who believe they can inuenceoutcomes at work are more likely to actually havean impact, and hence be more eective. Impactresults in more eort and greater persistence inthe face of obstacles (Abramson, Seligman, &Teasdale, 1978; Ashforth, 1989; Spreitzer et al.,1997; Thomas & Velthouse, 1990). Spreitzeret al. (1997) and Liden et al. (2000) found thatimpact was positively associated with work perfor-mance. These arguments and evidence lead to H4:

H4: There is a positive relation between the fourdimensions of psychological empowerment andmanagerial performance.

Role clarity and psychological empowerment

Finally, drawing on prior results, I hypothesizea positive relation between role clarity and psycho-logical empowerment. Unless SBU managers havea clear sense of their responsibilities and how toachieve them, it will be dicult for them to knowif they have the necessary skills and abilities to per-form their tasks adequately (i.e., feel empowered).As such, role clarity is expected to increase eachdimension of psychological empowerment; mean-ing, competence, self-determination and impact.Spreitzer (1996) argues that it is only when individ-uals understand their roles that those roles cantake on personal meaning. Clear lines of responsi-bility and clear task requirements are related tocompetence (Conger & Kanungo, 1988; Gist &Mitchell, 1992; Kahn et al., 1964). SBU managerswith clear work goals, and an understanding ofhow to achieve those goals, are likely to believethey can perform their job with skill and thus feelmore competent. Managers who are uncertain oftheir role expectations are likely to hesitate andnot take the initiative due to uncertainty, and thusexperience lower levels of self-determination (Spre-

148 M. Hall / Accounting, Organizaitzer et al., 1997). High levels of role clarity enableSBU managers of Australian manufacturing rmsfrom a commercial mailing list provider. Due tocost constraints, 400 managers were selected toform the sampling frame for the study. I used afour-step implementation strategy following therecommendations of Dillman (2000); telephonecalls to check data accuracy3, a questionnairepackage with cover letter, questionnaire andreply-paid envelope, a reminder postcard (senttwo weeks after questionnaire package), and a fol-low-up phone call (made two weeks after thereminder postcard). To encourage completion ofthe questionnaire, participants were promised asummary of the results and informed that their

3 The contact details of 31 of the 400 SBU managers couldnot be conrmed because they had ceased employment with thecontact organisation, the phone number was disconnected ordid not answer, or the organisation had ceased operations. As(Spreitzer et al., 1997). In contrast, individualswho understand their work roles are more likelyto take actions and decisions that inuence resultsin their work area (Sawyer, 1992). Prior researchshows that higher levels of role ambiguity arerelated to lower levels of psychological empower-ment (Smith & Langeld-Smith, 2003; Spreitzer,1996). This analysis indicates that role clarity willincrease each dimension of psychological empow-erment, which leads to H5:

H5: There is a positive relation between roleclarity and the four dimensions of psychologicalempowerment.

Research method

Sample selection and data collection

I collected data using a questionnaire adminis-such, the questionnaire was sent to 369 SBU managers.

responses were anonymous. Participants were alsoprovided with a practitioner article on PMS as atoken incentive (Davila, 2000; Dillman, 2000).

Of the 369 distributed questionnaires, 83 werereceived, which provides a response rate of22.5%.4 The response rate is similar to thosereported in recent accounting (Baines & Lang-eld-Smith, 2003; Moores & Yuen, 2001) andnon-accounting (Gordon & Sohal, 2001; Samson &Terziovski, 1999; Terziovski & Sohal, 2000) sur-veys of SBU managers in Australian manufactur-ing organizations. Due to the relatively lowresponse rate, I investigate the possibility of non-response bias. First, I compared the SBU size

pany policy not to respond to voluntary surveys,which are similar to the reasons for non-responsereported in other studies (for example, Baines &Langeld-Smith, 2003; Chenhall, 2005; Subraman-iam & Mia, 2003). These tests indicate that there isno signicant non-response bias in the sample.

Demographic information was collected fromrespondents regarding job tenure, company ten-ure, age, gender, SBU size (number of employees),and main manufacturing industry. Table 1 reportsthe descriptive statistics for the demographic vari-ables. The average age of respondents was 46.84years with an average job tenure of 5.14 yearsand an average company tenure of 10.64 years.Average SBU size was 336.13 employees. Eighty-two respondents were male and one was female.Table 2 reports the manufacturing industry classi-cation of respondents SBUs.

Table 1Descriptive statistics for demographic variables

Variable Minimum Maximum Mean St Dev

Job tenure(years)

1 25 5.14 5.95

Company tenure(years)

1 37 10.64 8.37

Age (years) 26 64 46.84 8.38SBU size (no.of employees)

10 4100 336.13 497.03

M. Hall / Accounting, Organizations and Society 33 (2008) 141163 149and industry representation of the 83 respondentsto the original list of 1000 SBUs. An independentsamples t-test shows that the mean sample SBUsize (X 336:13) is not signicantly dierent fromthe mean original list SBU size (X 566:93)(t = 1.400, p > 0.10). Furthermore, a v2-test showsthat the proportion of SBUs in each industry cat-egory is not signicantly dierent between thesample SBUs and original list SBUs (v2 = 5.981,degrees of freedom = 8, p > 0.10). Second, I com-pared early respondents (rst 20%) to late respon-dents (last 20%) for all constructs of interest(demographic and model variables). Results (notreported) show that there are no signicant dier-ences for any variables. In addition, during the fol-low-up phone calls I discussed with approximately40 non-respondents their reason(s) for not com-pleting the questionnaire. These reasons were timepressures, receiving too many surveys, and com-

4 16 cases contained missing data: 14 cases with one itemmissing, one case with two items missing, and one case withfour items missing. Littles MCAR test revealed that themissing data were missing completely at random (MCAR)(v2 = 4.424, degrees of freedom = 516, p > 0.10). As the missingdata is MCAR, any imputation method can be used (Hair,Anderson, Tatham, & Black, 1998). As such, the data werereplaced using the expectationmaximisation (EM) method inSPSS. The EM approach is an iterative two-stage process wherethe E-stage makes the best estimates of the missing data and theM-stage makes parameter estimates assuming the missing dataare replaced. This occurs in an iterative process until thechanges in the estimated parameters are negligible and themissing values are replaced (Hair et al., 1998; Little & Rubin,1987). This process resulted in a complete data set of 83

responses.n = 83.

Table 2Manufacturing industry classication

ANZSICa manufacturing industryclassication

Frequency %

21 Food, beverage and tobacco 8 9.6422 Textile, clothing, footwear andleather

3 3.61

23 Wood and paper products 6 7.2324 Printing, publishing and recordedmedia

3 3.61

25 Petroleum, coal, chemical andassociated products

12 14.46

26 Non-metallic mineral products 4 4.8227 Metal products 11 13.2528 Machinery and equipment 25 30.1229 Other 11 13.25

Total sample 83 100

a ANZSIC Australia and New Zealand Standard Industrial

Classication.

on the role of their PMS in providing performanceinformation. For all nine items, respondents wereasked to indicate on a 7-point Likert scale(1 = not at all to 7 = to a great extent) the extentto which each characteristic was provided by theirbusiness units PMS. The Appendix providesdetails of the explanatory statement and lists theitems in the scale.

Because the scale has not been used in priorresearch, I performed several tests to examine itspsychometric properties prior to including thescale in the PLS measurement model. As reportedin Table 3, the results of an exploratory factoranalysis show that the nine-item scale is unidimen-sional, with each item loading on the single factorabove 0.70. The Cronbach alpha for the nine-item

tions and Society 33 (2008) 141163Variable measurement

The questionnaire obtained information oncomprehensive PMS, psychological empowerment,role clarity and managerial performance. Estab-lished scales were used for each variable, exceptcomprehensive PMS. The development of thequestionnaire involved a review by three seniormanagement accounting academics with experi-ence in survey design. I also pilot tested the ques-tionnaire with four SBU managers (not part ofthe sample), who completed the questionnaireand participated in a brief interview. The reviewprocess and the pilot test resulted in minor changesto the wording of some items and to the layout ofthe questionnaire.

Ittner, Larcker, and Randall (2003) argue thatimproved methods are needed for determiningwhat rms mean by contemporary PMS, such asthe balanced scorecard. Prior research relating tocomprehensive PMS has used scales that examinethe extent to which a PMS contains a series ofspecic performance measures (for example,Hoque & James, 2000). A limitation of this typeof instrument is that it assumes that the perfor-mance measures contained in the instrument arerepresentative of the specic types of performancemeasures used by the rms in the sample. Firmsmay use similar nancial performance measures;however, non-nancial and/or strategic measuresare likely to be unique to each rm (Lipe & Salte-rio, 2000). In addition, this type of scale may notcapture the strategic linkages of more comprehen-sive PMS (Hoque & James, 2000). As such, Ideveloped a new scale to capture the comprehen-sive PMS construct. The scale consists of nineitems. Five items relate to the extent to which thePMS provides a variety of performance informa-tion about the important parts of the SBUs oper-ations. The remaining four items were drawn fromChenhall (2005), and relate to the extent of inte-gration of measures with strategy and across thevalue chain. The explanatory statement indicatedthat we were interested in the extent to which thePMS provides information about the operationsof the respondents business unit. This was doneto help ensure that when SBU managers were

150 M. Hall / Accounting, Organizaresponding to the statements, they were focusedscale is 0.95; well above acceptable limits (Nunally,1978). I also examined the extent to which the scaleconverged with an alternative measure of the com-prehensive PMS construct. Respondents were pro-vided with two descriptions of a PMS (reproducedin the Appendix). The rst description related to acomprehensive PMS (coded 1); the seconddescription related to a partial or less compre-hensive PMS (coded 0). Respondents indicatedwhich of the two descriptions better representedtheir PMS. The use of a forced-choice responseformat is consistent with the principle of usingmaximally-dissimilar forms of ratings when assess-ing convergent validity (Campbell & Fiske, 1959;Murphy & Davidshofer, 1998). The point-biserial

Table 3Factor loadings for nine-item comprehensive performancemeasurement system (CPMS) scale from an exploratory factoranalysis

Item Factor loading

CPMS1 0.915CPMS2 0.782CPMS3 0.843CPMS4 0.817CPMS5 0.896CPMS6 0.864CPMS7 0.852CPMS8 0.739CPMS9 0.836

Eigenvalue 6.350% Variance explained 70.559%n = 83.

correlation between the nine-item scale and the age or above average on each item. The Mahoneyet al. (1965) scale is frequently used to measuremanagerial performance in accounting studies(Chalos & Poon, 2000; Chong & Chong, 2002;Marginson & Ogden, 2005; Otley & Pollanen,2000; Parker & Kyj, 2006; Wentzel, 2002). Severalresearchers argue that self-report measures of per-formance are valid and tend to exhibit less bias

M. Hall / Accounting, Organizations and Society 33 (2008) 141163 151forced-choice scale is 0.66 (p < 0.001), which pro-vides strong support for the convergent validityof the nine-item scale.5 In addition, an indepen-dent samples t-test shows that the mean score onthe nine-item scale is signicantly higher for thoserespondents who indicated a comprehensivePMS (X 5:507) compared to those respondentswho indicated a partial PMS (X 3:827)(t = 7.867, p < 0.001). This supports the ability ofthe nine-item scale to distinguish between moreand less comprehensive PMS. The reliability andvalidity of the comprehensive PMS scale isassessed further in the PLS measurement model.

Established scales are used to measure role clar-ity, psychological empowerment, and managerialperformance, with their psychometric propertiesassessed in the PLS measurement model. Goalclarity and process clarity are measured with twove-item scales drawn from Sawyer (1992).Respondents were asked to indicate on a 7-pointLikert scale (1 = very uncertain to 7 = very cer-tain) the extent to which they were certain oruncertain about aspects of their job.

Psychological empowerment is measured withSpreitzers (1995) 12-item scale, with three itemsfor each empowerment dimension: meaning, com-petence, self-determination and impact. Respon-dents were asked to indicate on a 7-point Likertscale (1 = strongly disagree to 7 = strongly agree)the extent to which they agreed or disagreed witheach item.

As respondents are anonymous, it is not possibleto obtain supervisor ratings of managers perfor-mance. As such, managerial performance is mea-sured by a self-rated nine-item scale developed byMahoney, Jerdee, and Carroll (1965). The scaleassesses managerial performance along eightdimensions related to planning, investigating, coor-dinating, evaluating, supervising, stang, negotiat-ing and representing, and also includes an overallassessment of performance. Respondents wereasked to indicate on a 7-point Likert scale (1 = wellbelow average to 7 = well above average) theextent to which their performance was below aver-

5 I calculated the score for each respondent on the nine-item

scale as an average of the nine items.than supervisor ratings (Dunk, 1993; Marginson& Ogden, 2005; Parker & Kyj, 2006). In addition,prior research indicates that self-rated subjectivemeasures of subordinate performance are highlycorrelated with superiors subjective ratings of sub-ordinate performance and objective measures ofsubordinate performance (Furnham & Stringeld,1994; Heneman, 1974; Venkatraman & Ramanu-jam, 1987). The reliability and validity of the scalesis examined in the PLS measurement model.

Partial Least Squares regression

I use PLS regression analysis to test the hypoth-eses in this study. PLS is a latent variable modellingtechnique that incorporates multiple dependentconstructs and explicitly recognises measurementerror (Fornell, 1982), and has been used in a num-ber of accounting studies (Anderson, Hesford, &Young, 2002; Chenhall, 2004, 2005; Ittner, Larc-ker, & Rajan, 1997; Vandenbosch, 1999). PLS isparticularly suited to this study because it makesminimal data assumptions and requires relativelysmall sample sizes (Wold, 1985).6

PLS comprises a measurement model and astructural model. The measurement model speci-es relations between observed items and latentvariables. The structural model species relationsbetween latent constructs. In PLS the measurementand structural models are estimated simultaneously

6 Mardias (1970) test of multivariate kurtosis revealed thatthe data are multivariate non-normal (t = 26.076, p < 0.001).However, unlike structural equation modeling techniques suchas LISREL, PLS does not require normally distributed data.Because PLS is a regression based technique, it requires tencases for the most complex regression (Chin, 1998; Van-denbosch, 1999). In this study, the most complex regression isthat with managerial performance as the dependent variable,with eight independent variables, suggesting a minimum sample

size of 80 cases.

(Barclay, Thompson, & Higgins, 1995). However,the PLS model is typically interpreted in twostages. First, the reliability and validity of the mea-surement model is assessed. Second, the structuralmodel is assessed (Barclay et al., 1995). Thisensures that the constructs measures are reliable

removed from the scale and not used in further

scores for each variable are above 0.80, which dem-

152 M. Hall / Accounting, Organizations and Society 33 (2008) 141163analysis. The factor loadings from the nal PLSmeasurement model are reported in Table 4.

I assess the reliability of each variable using For-nell and Larckers (1981) measure of compositereliability and Cronbachs (1951) alpha. As shownin Table 5, the composite reliability and alpha

7 I obtained the PLS results using PLS Graph Version 3.0.8 An exploratory factor analysis (oblique rotation) of the

managerial performance scale shows two factors with eigen-values greater than one, with items MP1MP6 and MP9loading on the rst factor, and items MP7 and MP8 loading ona second factor. Thus, the low factor loadings for MP7 andMP8 arise because they do not form part of a unidimensionaland valid before assessing the nature of the rela-tions between the constructs (Barclay et al., 1995;Hair et al., 1998; Hulland, 1999). As such, theresults from the measurement model are presentedrst followed by an examination of the hypothes-ised relations between the constructs.7

Results

Measurement model

Statistics from the PLS measurement model areused to examine the psychometric properties of thevariables. First I examine the factor loadings foreach variable. All items load on their respectiveconstructs; however, two items from the manage-rial performance scale have factor loadings below0.5 (Hulland, 1999) (item MP7 = 0.461 and itemMP8 = 0.246). Low item loadings add very littleto the explanatory power of the model whilepotentially biasing the estimates of the parameterslinking the constructs (Chin, 1998; Hulland, 1999).Further tests show that the reason the two itemshave low factor loadings is because they do notform part of a unidimensional managerial perfor-mance scale.8 As such, items MP7 and MP8 aremanagerial performance scale (Barclay et al., 1995).onstrates acceptable reliability (Nunally, 1978).Convergent validity of the variables is assessed

by examining the average variance extracted(AVE) statistics. Table 5 shows that the AVE foreach variable is 0.50 and above, which demon-strates adequate convergent validity (Chin, 1998;Hair et al., 1998).

The AVE statistic is also used to assess discrim-inant validity by comparing the square root of theAVE statistics to the correlations among the latentvariables (Chin, 1998). This tests whether a con-struct shares more variance with its measures thanit shares with other constructs (Fornell & Larcker,1981). Table 5 shows that the square roots of theAVEs (diagonal) are all greater than the respectivecorrelations between constructs. In addition, Table4 shows that each item loads higher on the con-struct it intends to measure than on any other con-struct (Barclay et al., 1995; Chin, 1998). The resultsof these two tests demonstrate adequate discrimi-nant validity. Overall, the results from the PLSmeasurement model indicate that each constructexhibits satisfactory reliability and validity.

Tests of hypotheses

I estimate a structural model in PLS to test thehypotheses. In addition to the hypothesized paths,I also include job tenure in the structural model tocontrol for the endogeneity concern that more ten-ured employees have access to more informationand also feel more psychologically empowered(Chenhall & Moers, in press). The objective ofPLS is to maximise variance explained rather thant, therefore prediction-orientated measures, suchas R2, are used to evaluate PLS models (Chin,1998). The R2 for each endogenous variable isshown in Table 6. PLS produces standardised bsfor each path coecient, which are interpreted inthe same way as in OLS regression. As PLS makesno distributional assumptions, bootstrapping (500samples with replacement) is used to evaluate thestatistical signicance of each path coecient(Chin, 1998).9

9 Statistical signicance is determined using the reported

original PLS estimates and bootstrapped standard errors.

MEA

0.497

tionsTable 4Factor loadings from nal PLS measurement model

Item CPMS GC PC

CPMS1 0.920 0.431 0.196

M. Hall / Accounting, OrganizaAlthough there is a positive correlation betweencomprehensive PMS and managerial performance(see Table 5), Table 6 shows that comprehensivePMS is not associated with managerial perfor-mance (b 0:030; t 0:298; p > 0:10. As

PMS2 0.786 0.316 0.188 0.368CPMS3 0.837 0.280 0.135 0.404CPMS4 0.810 0.362 0.098 0.360CPMS5 0.896 0.365 0.190 0.352CPMS6 0.860 0.396 0.174 0.430CPMS7 0.859 0.380 0.134 0.435CPMS8 0.733 0.381 0.096 0.340CPMS9 0.841 0.361 0.142 0.398

GC1 0.349 0.787 0.532 0.450GC2 0.363 0.795 0.615 0.454GC3 0.265 0.786 0.530 0.426GC4 0.364 0.801 0.507 0.535GC5 0.399 0.875 0.566 0.535

PC1 0.203 0.586 0.795 0.295PC2 0.122 0.535 0.813 0.247PC3 0.115 0.590 0.847 0.363PC4 0.100 0.453 0.811 0.268PC5 0.185 0.596 0.817 0.450

MEAN1 0.457 0.559 0.366 0.949MEAN2 0.486 0.590 0.401 0.961MEAN3 0.405 0.534 0.375 0.918

COMP1 0.288 0.428 0.432 0.514COMP2 0.233 0.440 0.483 0.522COMP3 0.370 0.548 0.525 0.642

IMP1 0.502 0.538 0.353 0.619IMP2 0.346 0.545 0.540 0.490IMP3 0.413 0.535 0.461 0.481

SD1 0.264 0.400 0.455 0.334SD2 0.319 0.424 0.489 0.395SD3 0.183 0.430 0.572 0.302

MP1 0.262 0.412 0.302 0.529MP2 0.220 0.370 0.308 0.312MP3 0.244 0.385 0.342 0.258MP4 0.187 0.402 0.213 0.333MP5 0.223 0.509 0.394 0.540MP6 0.265 0.532 0.413 0.456MP9 0.170 0.511 0.449 0.500

CPMS, comprehensive performance measurement system; GC, goaltence; IMP, impact; SD, self-determination; MP, managerial performn = 83.N COMP IMP SD MP

0.407 0.475 0.269 0.319

and Society 33 (2008) 141163 153expected, this indicates that comprehensive PMSdoes not have a direct eect on managerial perfor-mance, but, instead, its eect on managerial per-formance is fully mediated by the interveningvariables. The results from the structural model,

0.318 0.419 0.226 0.2360.300 0.308 0.159 0.3050.158 0.382 0.201 0.2010.308 0.426 0.249 0.2380.261 0.379 0.281 0.3160.315 0.491 0.268 0.2510.159 0.307 0.175 0.2510.365 0.457 0.268 0.254

0.431 0.564 0.478 0.4810.498 0.538 0.473 0.5590.335 0.378 0.357 0.4300.457 0.517 0.244 0.5800.488 0.483 0.326 0.524

0.446 0.464 0.448 0.4480.488 0.408 0.501 0.3360.426 0.394 0.427 0.4140.357 0.339 0.455 0.3640.533 0.467 0.481 0.462

0.593 0.556 0.361 0.5830.632 0.648 0.407 0.5970.625 0.525 0.305 0.554

0.868 0.453 0.336 0.4710.777 0.566 0.403 0.3300.932 0.560 0.473 0.559

0.618 0.850 0.495 0.5190.480 0.881 0.644 0.3320.478 0.886 0.683 0.366

0.318 0.634 0.891 0.2480.474 0.657 0.938 0.3820.461 0.580 0.871 0.418

0.431 0.337 0.228 0.6640.294 0.276 0.236 0.5680.246 0.237 0.291 0.7010.280 0.289 0.138 0.6890.411 0.357 0.350 0.7680.460 0.372 0.298 0.7430.450 0.415 0.372 0.788

clarity; PC, process clarity; MEAN, meaning; COMP, compe-ance.

reported in Table 6, indicate how role clarity andpsychological empowerment act as interveningvariables in the relation between comprehensivePMS and managerial performance.

For the role clarity path, there is a positive asso-ciation between comprehensive PMS and goalclarity (b = 0.440, t = 4.668, p < 0.01), and a weakpositive association between comprehensive PMS

Table5

Descriptive

statistics,reliabilityandaveragevariance

extracted(AVE)statistics,andcorrelationsfrom

PLSmodel

Variable

Mean

Standarddeviation

Cronbachalpha

Compositereliability

AVE

Correlations

CPMS

GC

PC

MEAN

COMP

IMP

SD

MP

CPMS

4.657

1.289

0.946

0.955

0.705

0.840

GC

5.963

0.829

0.868

0.905

0.655

0.434

0.809

PC

5.191

0.871

0.875

0.909

0.667

0.182

0.681

0.817

MEAN

5.916

1.019

0.935

0.960

0.889

0.478

0.595

0.404

0.943

COMP

5.891

0.729

0.804

0.895

0.742

0.351

0.552

0.557

0.654

0.861

IMP

6.121

0.778

0.842

0.905

0.761

0.487

0.619

0.513

0.613

0.608

0.872

SD

5.971

0.938

0.880

0.929

0.811

0.281

0.466

0.567

0.381

0.472

0.691

0.902

MP

5.405

0.610

0.824

0.874

0.500

0.315

0.641

0.500

0.613

0.536

0.472

0.397

0.707

CPMS,comprehensive

perform

ance

measurementsystem

;GC,goal

clarity;

PC,process

clarity;

MEAN,meaning;

COMP,competence;IM

P,impact;SD,self-

determination;MP,managerialperform

ance.

Diagonalelem

entsarethesquarerootsoftheAVEstatistics.O-diagonalelem

entsarethecorrelationsbetweenthelatentvariablescalculatedinPLS.Allcorrelations

above

0.20

arestatisticallysignicant(p 0.10).

eects of management control systems on work

Table6

PLSstructuralmodel:pathcoe

cients,t-statistics

andR2

Dependentvariables

Independentvariables

Comprehensive

PMS

Goalclarity

Processclarity

Meaning

Competence

Self-determination

Impact

Jobtenure

R2

Comprehensive

PMS

0.121(1.365)*

0.016

Goalclarity

0.440

(4.668)***

0.050

(0.600)

0.200

Processclarity

0.184

(1.497)*

0.020

(0.237)

0.034

Meaning

0.256

(2.845)***

0.448

(3.153)***

0.052

(0.445)

0.143(2.100)**

0.442

Competence

0.167

(1.474)*

0.224

(1.594)*

0.374

(3.369)***

0.151(1.857)**

0.417

Self-determination

0.147

1.305)*

0.064

(0.492)

0.495

(4.171)***

0.143(1.647)**

0.375

Impact

0.292

(2.717)***

0.334

(2.607)***

0.233

(2.328)***

0.063(0.835)

0.480

Managerialperform

ance

0.030

(0.298)

0.380

(2.328)***

0.047

(0.311)

0.351

(2.350)***

0.113

(0.826)

0.107

(0.833)

0.129

(0.591)

0.027

(0.379)

0.513

n=83.Eachcellreportsthepathcoe

cient(t-value).Blankcellsindicatethat

thepathwas

nothypothesized

within

themodel.

*p