shields deng kato2000

8/4/2019 Shields Deng Kato2000

1/18

The design and eects of control systems: tests of direct- and indirect-eects models

Michael D. Shields a, *, F. Johnny Deng b , Yutaka Kato ca Department of Accounting, Broad Graduate School of Management; Michigan State University,

East Lansing, MI 48824-1121, USAb University of Arkansas, USA

cKobe University, Japan

Abstract

Two models are developed on the eects of a control system that include participative standard setting, standard-based incentives, and standard tightness. The direct model proposes that the control system directly aects perfor-mance, whereas the indirect model proposes that the eects of the control system on performance are indirect throughthe mediating inuence of job-related stress. Hypothesis tests and a models-comparison test using structural equationmodeling indicate that the indirect model has a signicantly better t to the data than does the direct model. # 2000Elsevier Science Ltd. All rights reserved.

Control systems as researched in the accountingliterature are usually based on a cybernetic modelin which performance standards (e.g. budget, goal,target) 1 and performance measures are comparedas the basis for corrective action and performanceevaluation. Three important components of per-formance-based control systems are the standardsetting process (e.g. participation, imposition),

standard tightness (e.g. goal di culty, budgetaryslack), and standard-based incentives (e.g. a bonusfor each unit of measured performance in excess of the performance standard). These three compo-nents are interrelated (Demski & Feltham, 1978).The process used to set a performance standard islikely to aect its tightness which, in turn, mayaect the rewards received for performance rela-tive to the standard.

Many studies have investigated the direct eects 2

of one or more components of control systems(e.g. budget participation, standard tightness, per-formance-based incentives) on performance orother variables (e.g. job-related stress) (Birnberg,Shields & Young, 1990; Kren & Liao, 1988;Merchant, 1989; Shields & Shields, 1998; Young,

0361-3682/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.P I I : S 0 3 6 1 - 3 6 8 2 ( 9 9 ) 0 0 0 4 1 - 0

Accounting, Organizations and Society 25 (2000) 185202www.elsevier.com/locate/aos

2 A direct eect is the inuence of one variable on anotherthat is not mediated by any other variable. An indirect eect isa direct eect that is mediated by at least one other variable.See Bollen (1989).

* Corresponding author. Fax:+1-517-432-2915.1 Hereafter, we use the terms standard and standard setting

instead of budget and budgeting although they have similarmeanings (Atkinson, 1978; Christensen, 1982; Horngren, Fos-ter & Datar, 1997; Merchant, 1982). Results of prior researchin accounting and organizational behavior and psychology inwhich job or task requirements are expressed as goals, budgets,standards, or targets are qualitatively similar [see the ensuingliterature review and Locke and Latham (1990) and Shields andShields (1998)]. For the sample used in this study, task goals arestated as standards, but this should not be confused with stan-dard costs (i.e. formulas for mapping resource costs to out-puts). Instead, as used in this paper, standard means the level of performance expected by an organization.


2/18

1988; Young & Lewis, 1995). These studies haveinvestigated either the independent direct eectsof a control-system component (e.g. bivariate

correlation between participative standard settingand performance) or the interactive direct eectsof a component (e.g. a component interacts withanother independent or a moderator variable toaect performance). Few studies have investigatedwhether a control-system component has indirecteects on performance [e.g. a model with anintervening variable(s) that mediates the relation-ship between the component and performance].

While these prior studies have provided manyinsights into the eects of control-system compo-nents, most studies have one or more of the follow-ing three limitations. First, many studies havelimited their investigation to one control-systemcomponent, while fewer studies have focused onmultiple components. Since management account-ing and controls usually have more than one com-ponent and are represented as a system, it isimportant for research to include multiple compo-nents in order to understand how they operate as asystem and the system's eects. There, however, is apractical limit to the number of control-systemcomponents that can be included in any one study

due to the need to have theoretical and predictiveclarity and an adequate sample size to test the pre-dictions. In addition, including many components inan empirical study requires estimating the magni-tude and direction of the numerous predicted cause-eect relations in the sample (i.e. relationshipsbetween components and relationships betweencomponents and other variables).

Second, while control systems and their compo-nents can have many eects (e.g. job-related stress,performance, conict, cooperation, communica-tion), most studies include only one dependentvariable or, if multiple dependent variables areincluded, relationships among them are not inves-tigated. 3 Limiting a study to one dependent vari-able or including multiple dependent variables but

suppressing relationships among them can lead to adistorted understanding of the eects of controlsystems. The predicted or estimated eects of a

control-system component on a dependent vari-able can depend on which other variables that areaected by that component are included in the study.As will be shown in this study, for example, thereported results and implications of a study candepend on whether it includes as consequences of acontrol system performance, job-related stress, orboth variables but either suppressing or incorporatingthe relationship between stress and performance.

Third, almost all of these studies investigate thedirect (independent or interactive), but not indir-ect, eects of control systems. There can be impor-tant theoretical dierences between direct andindirect models that have practical implications.For example, Bollen (1989, p. 376) suggests that,

The indirect and total eects can help toanswer important questions that are notaddressed by examining the direct eects. Forinstance, suppose that participation in asocial welfare program has a positive directeect on household income but a negativeindirect eect because it reduces the number

of hours of part-time work, which in turnreduces other sources of income. The directeect provides a misleading impression of theinuence of program participation on income.

There is a similar dierence between the predic-tions of the direct- and indirect-eects models tes-ted in this paper: they imply opposite managementpolicies for the level at which performance stan-dards should be set to increase or maximize per-formance.

The purpose of this paper is to develop and testtwo competing control system models. Each hasthree control-system components participativestandard setting, standard-based incentives, andstandard tightness and job performance as thedependent variable. The direct model, based onprior accounting research, assumes that the threecontrol-system components directly aect job per-formance. The indirect model, based on organiza-tional psychology literature that did not inuencethe prior accounting studies, assumes that the eects

3 A dependent variable is directly caused by an independentvariable. An intervening variable is caused by an independentvariable and causes a dependent variable. A moderator variableaects the relationship between an independent and dependentvariable but does not have a bi-variate relationship with eitherthe dependent or independent variable.

186 M.D. Shields et al. / Accounting, Organizations and Society 25 (2000) 185202


3/18

of the control-system components on job perfor-mance are indirect, through the intervening variableof job-related stress. Structural equation modeling is

used to test the proposed models and relatedhypotheses. The results indicate that the indirectmodel provides a better t to the data than does thedirect model; all of the hypotheses that constitutethe indirect model are supported, but two of vehypotheses that comprise the direct model are not.

The remainder of this paper is organized in foursections. The next section provides a review of prior accounting and organizational psychologystudies as the basis for developing two competingcontrol-system models and the nine hypothesesthat in total constitute the two models. The nexttwo sections describe the hypothesis testingmethod and results. The last section discusses theresults and their implications for future research.

1. Literature review and hypothesis development

An important determinant of the expected designand eects of a control system is the nature of the job or task being controlled. Appendix A provides adescription of the job/task. The job is automobile

design engineering and the task is designing a parti-cular component(s) of an automobile. The designengineers are Japanese nationals who are employedby a large Japanese company that sells automobilesaround the world. They work at this company'sfacility in Japan where automobiles are designed.This company uses a target costing system whichhas tight performance standards. The performancemeasures used for these employees and their jobs arespecic to each job and task (i.e. qualitative, quanti-tative, nancial, nonnancial). 4 Three components of the control system used for this job are participativestandard setting, standard tightness, and standard-based incentives. The eects of these control-systemcomponents are on job-related stress and jobperformance.

Numerous accounting studies have investigatedthe eects of participative standard setting, stan-dard tightness, or standard-based incentives. Per-

formance is the most frequently researched (andprobably the most important) eect of these con-trol-system components, and job-related stress hasbeen a dependent variable in some studies. Thesestudies have also investigated the relationshipbetween participative standard setting and bothstandard tightness and standard-based incentives.

Two competing models of the eects of this three-component control system are proposed and tested.Each model in Fig. 1 includes H2 and H3, whichdescribe the theoretical relationships between parti-cipative standard setting and both standard tight-ness and standard-based incentives. The modelsdier in their predictions about the eects of thiscontrol system. The direct model is based on theextant accounting research (to be reviewed later)and predicts that performance is directly caused byeach control-system component (i.e. H1, 4, 5 in Fig.1). In contrast, the indirect model predicts that theeects of the control system on performance areindirect through the mediating inuence of job-related stress as the intervening variable (i.e. H69 inFig. 1). These models have dierent implications for

how to improve performance. The direct modelimplies that expected performance can be increasedif the performance standard is tightened. In con-trast, the indirect model predicts the opposite loosening the performance standard decreasesexpected job-related stress which, in turn, increasesexpected performance. 5 The remainder of this sec-tion rst reviews literature to develop the hypothesesthat constitute the direct model, and then those thatmake up the indirect model.

1.1. Participative standard setting

Subordinate participation in setting performancegoals, standards, or budgets is one of the mostresearched topics in management and accounting

4 The company's reliance on several performance measures,in particular nonnancial qualitative and quantitative measuresthat are tailored to various jobs/tasks, is consistent with theevidence in Hertenstein and Platt (1998) and Nixon (1998)about the performance measures used to monitor and evaluatenew product development activities and employees.

5 There, however, are practical limits to how loose (slack) aperformance standard can be (e.g. the minimum required per-formance, resource availability constraints). Loose, or the loos-ening of, performance standards is consistent with evidence thatmany rms intentionally use highly-achievable budgets (e.g.achieved 8090% of the time) (Merchant & Manzoni, 1989).

M.D. Shields et al. / Accounting, Organizations and Society 25 (2000) 185202 187


4/18

(Locke & Latham, 1990; Shields & Shields, 1998). Itcan be used by superiors and subordinates to deter-mine the level or tightness of the standard and thereward for performance relative to the standard.

For example, Shields and Waller (1988) had a par-ticipative standard setting situation in which super-iors designed menus of contracts to oersubordinates and then the subordinates selected thecontracts under which they would work and be paid.The contracts diered in terms of the xed salary,the bonus per unit that actual performance exceedsthe standard performance, and the level of the per-formance standard. A commonly used expression of the relationship among the performance standard,standard-based incentives, and performance in theaccounting literature is (Young & Lewis, 1995):

r b a s if a b sY

0 if a ` sY&

where,

r rewardb bonus per unita actual performances performance standard.

When the parameters of this formula are deter-mined during participative budgeting, a subordinateand superior can have dierent preferences forthem. When a subordinate participates, he or she is

assumed to try to maximize r by negotiating for anoptimized combination of b and s (e.g. large b andlow s). The superior is assumed to use participativebudgeting to learn more about the subordinate'stask in order to set a b s pair that maximizes a r.

1.2. Participative standard setting and job performance

The direct eects of participative standard set-ting on job performance has been investigated innumerous studies (Shields & Shields, 1998). 6 Theearliest studies relied on organizational and psy-chological theories to structure their investigationof the independent direct eects of participation on

Fig. 1. Theoretical model.

6 Many participation studies in the accounting literature arebased on research in organizational behavior and psychology onparticipative goal setting, decision making, and management(Locke & Latham, 1990). The qualitative results of the account-ing and non-accounting studies are similar although a higherpercentage of the accounting studies report detecting statisticallysignicant direct eects of participation on performance.



5/18

attitude, job-related stress, motivation, perfor-mance, and satisfaction (e.g. Hofstede, 1967;Kenis, 1979; Milani, 1975). The evidence from

these studies is somewhat mixed and weak, but themajority supports a positive, direct relationship.Hopwood (1976) notes that these studies and rela-ted organizational studies usually report that par-ticipation has positive, independent, direct eectson attitude and satisfaction, but the evidence onperformance is equivocal at best. In particular,some studies report positive, independent, directeects of participation on performance under cer-tain circumstances while other studies report nega-tive, independent, direct eects for other situations.

In response to Hopwood's (1976) interpretationof the extant literature, studies began to try to sortout the mixed and weak results of the priorresearch by hypothesizing, based on a variety of organizational and psychological theories, thatparticipation as an independent variable interactswith another independent or a moderator variableto directly aect performance and other variables(e.g. Govindarajan, 1986; Merchant, 1981). Themajority of the evidence in these studies indicatesthat participation is part of an interaction (withanother independent or a moderator variable) that

has a positive, ordinal-interaction, direct eect on job performance (Shields & Shields, 1998). Almostall of the studies have investigated the independentor interactive direct eects of participation andvery few have examined its indirect eects. 7 Intotal, most studies investigating the independentor interaction direct eects of participative stan-dard setting report a positive, direct relationshipbetween it and job performance.

H1. There is a positive relationship between partici-pative standard setting and job performance.

1.3. Participative standard setting and standard tightness

Theoretical analysis and eld-study and experi-mental evidence indicates that a negative relation-ship between participative standard setting andstandard tightness should be expected. Thisexpected relationship stems from theoreticalresearch in organizational behavior and econom-ics which predicts that allowing subordinate par-ticipation in setting performance standards canresult in slack. Slack arises during participativestandard setting when a subordinate intentionallyunderstates the expected revenues and/or produc-tive capability, and/or overstates the expectedcosts and/or resources required to achieve a per-formance standard. 8 A consequence of participa-tive standard setting, therefore, is thatperformance standards may not be as tight as theywould have been if there were no participation.

The initial research on participation and slack,based on Cyert and March (1963) and Williamson(1964), argued that building slack into a standard is

in a subordinate's best economic interest, and pro-vided eld-study evidence which showed that slackexists and identied factors that might be related toit (Lowe & Shaw, 1968; Schi & Lewin, 1970). Sub-sequent studies suggested that participation reducesa subordinate's need to build slack into the budget;but this negative relationship between participationand slack was only indirectly detected by Onsi(1973), not detected by Collins (1978), and veryweakly detected by Merchant (1985).

More rigorous predictions were subsequentlyderived based on the economic theory of agency(Baiman, 1982; Baiman & Evans, 1983). Thebasic prediction is a positive relationship betweenparticipation and slack because, when a sub-ordinate's expected compensation is related tostandard tightness, ceteris paribus, he or she isexpected to use the opportunity to participate instandard setting to build slack into the standard in

7 Two studies investigate indirect eects. Brownell andMcInnes (1986) report, contrary to their prediction, that parti-cipative budgeting has a nonsignicant indirect eect on per-formance through motivation as the intervening variable (thisindirect path was not signicant because the participation-motivation linkage was not signicant), and a signicant inde-pendent, direct eect on performance. Nouri and Parker (1998)predict and nd that the inuence of participative budgeting onperformance is indirect, with separate paths through organiza-tional commitment and budget tightness as intervening vari-ables. All of the linkages in these studies' models are positive insign which means that the indirect eects would be expected toaect the strength, but not the sign, of the expected direct eectof the independent variable on the dependent variable.

8 See Dunk and Nouri (1998) for similar denitions of slackand a review of research on slack.



6/18

order to increase expected compensation. Con-sistently, Lukka (1988) provides eld-study evi-dence that when a subordinate participates in

setting his or her budget and rewards are tied toperformance relative to the budget, he or she triesto use participation to get an easier budget.

Several studies test the participation-tightness/slack relationship based on this agency-theoryprediction. Nouri and Parker (1998) report surveyevidence that participative budgeting is related tolooser budgets. Experimental studies report thatsubordinates who participate by ``self selecting'' theirown performance standard choose lower standards(Chow, Cooper & Haddad, 1991; Chow, Cooper &Waller, 1988; Waller, 1988; Young, 1985). Overall,the theoretical analysis and experimental, survey,and eld-study evidence supports the hypothesis of a negative relationship between participative stan-dard setting and standard tightness.

H2. There is a negative relationship between par-ticipative standard setting and standard tightness.

1.4. Participative standard setting and standard-based incentives

Prior research has predicted based on agencytheory that, when participative standard setting isused, management will make greater use of incen-tives that reward performance relative to meetingor exceeding the standard (Baiman, 1982; Baiman& Evans, 1983). The assumption underpinningthis relationship is that as the level of participationincreases, a superior can learn how to develop abetter performance standard, one that will moti-vate a subordinate to maximize performance(Shields & Shields, 1998). An important way tomotivate the subordinate to maximize perfor-mance is to provide more incentive to achieve andsurpass the performance standard, for example byincreasing the unit bonus for performance inexcess of the standard.

We also expect that a subordinate will use theopportunity to participate to increase standard-based incentives. The subordinate is assumed to beusing participation to obtain the combination of standard-based incentives and standard tightnessthat maximize his or her expected reward for

example, a particular high level of unit bonus forperformance over the standard and a particularlow level of performance standard. That is, ceteris

paribus, we would expect a subordinate to useparticipative standard setting to try to increase his orher expected reward for the excess of measured per-formance over the standard. Shields and Young(1993), the only evidence on this issue, report a posi-tive relationship between participative budgetingand budget-based incentives.

H3. There is a positive relationship between parti-cipation and standard-based incentives.

1.5. Standard tightness and job performance

Standard tightness is dened as the amount of resources needed to perform at the level of a stan-dard minus the amount of resources provided toperform. Several theories of motivation, includinglevel of aspiration (Stedry, 1960), expectancy(Ronen & Livingstone, 1975), agency (Chow,1983), and goal setting (Hirst & Lowy, 1990),have been used to predict that, assuming a stan-dard is accepted as attainable, performance is anincreasing function of standard di culty. Con-sistent with these theoretical predictions and theresults of many studies in organizational psychol-ogy (Locke & Latham, 1990), experimental andsurvey research in accounting report a positiverelationship between standard tightness and indi-viduals' standard-related performance (Chow;Hirst & Lowy; Hofstede, 1967; Lindquist, 1995;Rockness, 1977; Stedry, 1960; Waller & Chow,1985). In contrast, Kenis' (1979) survey evidence isambiguous with respect to the relationship betweentightness and performance. He reports that whiletightness is positively related to motivation, it isnegatively related to performance. However, thislatter result depends on the perceived tightness (per-formance decreases when standards are perceived tobe tighter than ``about right''). Simons (1988)extends the evidence from individual to organiza-tional performance and reports, consistent withalmost all of the individual-level evidence, that tigh-ter budget goals are associated with higher levels of organizational performance. Overall, almost all of the evidence from the prior research supports the



7/18

hypothesis of a positive relationship betweenstandard tightness and job performance.

H4. There is a positive relationship between stan-dard tightness and job performance.

1.6. Standard-based incentives and job performance

Agency, expectancy, and goal-setting theorieseach predict that performance is an increasing func-tion of performance-contingent incentives (Demski &Feltham, 1978; Locke & Latham, 1990; Ronen &Livingstone, 1975). The essence of this theor-eticalprediction is that individuals are motivated to expendmore eort when they believe that the additionaleort will increase performance which, in turn, resultsin receiving additional valued rewards. Specically,incentives contingent on the excess of performanceover a standard will motivate an individual to exertadditional eort up to the point that the cost of theincremental eort equals the benet received fromthat eort (e.g. increase in bonus). The accountingresearch that is based on predictions from thesetheories report evidence of a positive relationshipbetween standard-based incentives and perfor-mance (Chow, 1983; Dillard & Fisher, 1990; Kren,

1990; Rockness, 1977; Waller & Chow, 1985).

H5. There is a positive relationship between stan-dard-based incentives and job performance.

1.7. Job-related stress

Early research on job-related stress is based onthe arousal model (a variant of the YerkesDodsonLaw), which predicts that stress increases motiva-tional arousal which, in turn, increases and focuseseort and hence performance. However, at some(moderate) stress level, additional arousal causesinterference, anxiety, or disorganization which redu-ces and diuses eort and hence performance. Thearousal model predicts an inverted- U relationshipbetween job-related stress and performance (Beehr,1985; McGrath, 1976; Weick, 1983); however, ingeneral the evidence did not support the prediction.

In response to the lack of predictive ability of this arousal model of stress, the task demand-per-formance capability model was developed based

on the person-environment t literature (Beehr1985; Beehr & Bhagat, 1985; Edwards, 1996; VanHarrison, 1985). This model predicts that, ceteris

paribus, job-related stress is a positive function of the dierence between the performance demandsof a task (i.e. the task goals and constraints) andthe individual's performance capability (e.g. skill,eort, external resources). As task demandsincrease over performance capability, job-relatedstress is assumed to increase due to either ambi-guity about task demands or task demandsexceeding performance capability (overload). Asdemand ambiguity or overload increase, an indi-vidual has more uncertainty about the possibleoutcomes of the job and/or how his or her eortaects job outcomes, resulting in feelings of roleambiguity and/or loss of control, which then cau-ses a reduction and diusion in eort and henceperformance. Several studies provide evidenceconsistent with the predicted determinants andeects of job-related stress (Beehr; Beehr &Bhagat; Edwards; Motowidlo, Pacard & Manning,1986; Van Harrison). The ensuing subsection ela-borate on how the control-system componentscause job-related stress and on the eects of stress.

1.8. Participative standard setting and job-related stress

The extant accounting research on the relation-ship between participation and job-related stress isweak on theory and has reported mixed results.Kenis (1979) did not explicitly predict, but reportsnding, a signicant negative relationship. Threestudies predict that participative budgeting is part of a particular interaction that causes job-relatedstress, but their results are contrary to prediction: (1)Brownell and Hirst (1986) found a negative linearrelation between participation and job-relatedstress; (2) Harrison (1992) reports that participationis part of a signicant interaction that is associatedwith job-related stress but that interaction is not thepredicted one, and the bivariate relationship is notsignicant; and (3) Lau, Low and Eggleton (1995)did not nd their predicted interaction, nor is thelinear bivariate relationship signicant.

Consistent with the results of Brownell and Hirst(1986) and Kenis (1979), but not their predictions, the



8/18

primary prediction and result of research based on thetask demands-performance capability model is thatparticipation in decision making causes job-related

stress to decrease (Beehr, 1985; Jex & Beehr, 1991).The root of this theoretical relationship is that parti-cipation increases individuals' feelingof control whichthereby reduces their stress. Thus, we propose that job-related stress decreases as participation increases.

H6. There is a negative relationship between par-ticipative standard setting and job-related stress.

1.9. Standard tightness and job-related stress

Theoretical research based on the task demands performance capability model predicts that, ceterisparibus, job-related stress is a positive function of the dierence between the performance demands of a job and individuals' performance capability, whichwe refer to as standard tightness (Beehr 1985; Beehr& Bhagat, 1985; Edwards, 1996; Van Harrison,1985). This model of job-related stress proposesthat stress increases as the demands of a jobincreasingly exceed a worker's performance cap-ability. Several studies provide evidence consistentwith this prediction (Beehr; Beehr & Bhagat;

Edwards; Motowidlo et al., 1986; Van Harrison). Ina budgetary context, Jick's (1984, 1985) analysissupports the prediction of a positive relationshipbetween budget tightness and job-related stress.Similarly, Kenis (1979) reports evidence of a posi-tive relationship between budget tightness and job-related stress. In the ensuing hypothesis, standardtightness refers to a standard that the subordinatebelieves is achievable for two reasons: (1) an unat-tainable standard would not likely result from par-ticipation; and (2) an unattainable standard is noteective in the long-term since the subordinate willnot be motivated to achieve a standard that he orshe consistently believes is unattainable.

H7. There is a positive relationship between stan-dard tightness and job-related stress.

1.10. Standard-based incentives and job-related stress

Prior accounting research has not addressed therelationship between standard-based incentives

and job-related stress. A negative relationshipbetween standard-based incentives and job-relatedstress is predicted based on prior stress research

that nds that when individuals expect higherrewards for achieving a goal, they will experi-ence less stress because the expected rewardsincrease performance capability by arousing andfocusing eort (Edwards, 1996; Van Harrison,1985). This increased eort makes the task lessdicult, thus reducing the excess of perfor-mance demands over performance capabilitywhich then reduces stress. 9

H8. There is a negative relationship between stan-dard-based incentives and job-related stress.

1.11. Job-related stress and job performance

As previously presented, the task demand-per-formance capability model predicts a negativerelationship between job-related stress and jobperformance (Beehr, 1985; Beehr & Bhagat, 1985;Edwards, 1996; Van Harrison, 1985). This modelassumes that stress results from either ambiguityabout task demands or task overload. As taskdemand ambiguity or overload increase, an indi-

vidual has more uncertainty about the possibleoutcomes of the job and/or how his or her eortaects job outcomes, resulting in feelings of roleambiguity and/or loss of control, which then causesa reduction and diusion in eort and hence per-formance. The evidence outside accounting(Jamal, 1985; McGrath, 1976; Motowidlo et al.,1986) and in accounting (Dunk, 1993) is consistentwith the prediction of a negative relationship.

H9. There is a negative relationship between job-related stress and job performance.

9 When there is a high level of incentives (in the extreme,``excess'' incentives over those needed to motivate achieving thestandard), the excess will not cause stress when it can be ignoredor retained by the performer for other purposes (e.g. leisure,engaging in job-related behavior which is not implied by theperformance standard). Only when these excess incentives hin-der or interfere with the achievement of the current performancestandard will they cause job stress (Edwards, 1996; Van Harri-son, 1985). The negative eect of incentives is not applicable forour sample because the excess incentives can be ignored or usedto motivate accomplishment of other valued goals.



9/18

2. Hypothesis-testing method

This section has three parts. The rst part fur-

ther describes the sample that is used to test thehypotheses, the second part describes how thequestionnaire that was used for data collectionwas developed, and the last part describes how thevariables were measured.

2.1. Sample

The sample was obtained by having a seniorexecutive of the automobile division of a globalJapanese company distribute the questionnaires tothe automobile design engineers who work in thisrm's design facility in Japan. The completedquestionnaires were mailed by these engineersdirectly to one of the authors. Of 480 ques-tionnaires distributed, 415 (86%) were returned.However, only 358 were usable because 46respondents' self-reported job titles were notdesign engineers and 11 had missing data. These358 usable subjects had a mean of 11.2 (SD=7.3,range=133) years of employment with thiscompany and a mean of 7.8 (SD=6.1, range=0 32) years of experience in their current job

assignment.10

2.2. Questionnaire

The questionnaire was initially developed inEnglish, then translated into Japanese and back-translated into English by dierent individuals.The original and back-translated versions werecompared. The Japanese version was revised basedon the detected dierences, then translated backinto English and again compared to the originalEnglish version. Only a few remaining dierenceshad to be resolved to nalize the Japanese version.

2.3. Measurement of variables

Each variable was measured using multipleindicators. Existing measures were used when

possible, with modication to t the presentresearch context. To check on the relevance of these measures to the subjects, one of the authors

who has had a consulting relationship with thiscompany showed the questionnaire to representa-tives of the company and they indicated that theybelieved the items in the questionnaire were rele-vant to the design engineers. The items used tomeasure each variable are in Appendix B.

Participative standard setting was measured byeight items adapted from Shields and Young(1993). Each item was measured on a seven-pointscale with higher values representing greater par-ticipation.

Standard tightness has been measured severalways in the prior research. We adopted a mea-surement approach that was consistent with ourdenition of standard tightness, which was alsoconsistent with how the stress literature con-ceptualized sources of stress. It was measured asthe dierence between two constructs resourcesneeded (as perceived by a respondent to accom-plish a job) and resources available (to accomplishthe job as perceived by a respondent). Theresponse to each of the two constructs for eachtype of resource was captured on a seven-point

response scale. Then for each of four types of resources (your own time, design technology, sup-port from other employees, support from outsidesources), resources needed minus resources avail-able was calculated. Each respondent's four dif-ferences between resources needed and availablewere then summed. The summed score was re-scaled so that it ranged from one to 13, withhigher values representing a tighter performancestandard.

Standard-based incentives were measured with amodied version of the three-item instrument usedby Shields and Young (1993). The scores of thethree seven-point incentive measures were summedwith higher values indicating more incentives.

Job-related stress was measured as the sum of responses to nine items from the 15-item instru-ment originally developed by Kahn, Wolfe,Quinn, Snoek and Rosenthal (1964) and has beenused in several accounting studies [e.g. Brownelland Hirst (1986)]. This instrument was modied tot the context of the present research, including

10 The hypothesis testing results did not qualitatively dierwhen those design engineers with less than three years of experience were deleted from the sample.



10/18

deleting six items which were not applicable. 11

Each item was measured on a ve-point scale,with higher values indicating higher stress.

Job performance of a design engineer was notmeasured by a nancial measure (e.g. prot, ROI).As discussed previously, nancial performancemeasures captured only part of a design engineer'smeasured performance. We could not use thecompany's performance evaluation that would beapplicable to each engineer because we weredenied access to them. Moreover, these perfor-mance measures and evaluations tend to be job,task, and department specic, which reducedcomparability across engineers. As an alternativeapproach to performance measurement, manycontrol system studies have used the Mahoney,Jerdee and Carroll (1965) instrument in whicheach respondent subjectively rated his or her ownperformance on eight dimensions of managerialactivity (e.g. planning, stang, supervising) as wellas on overall managerial performance. Thisapproach to data collection was feasible, thoughthe instrument was not appropriate for measuringthe performance of design engineers since they donot primarily perform managerial activities. Thus,we developed a measure in which each respondent

subjectively rated his or her performance on threeitems, each with a seven-point scale, with highervalues indicating higher performance. 12

3. Results

This section has four parts. The rst part reports

descriptive statistics for the variables, the secondpart describes the structural equation model, andthe other two parts present the results of themodels-comparison tests and hypothesis tests.

3.1. Descriptive statistics

Table 1 reports the variables' Cronbach alpha,means, standard deviations, and theoretical andactual ranges. Each variable's Cronbach alphawas greater than 0.60 which indicated that theyhad at least a satisfactory level of univariate relia-bility. The variables' actual ranges were almost aslarge as their theoretical ranges and their meanswere approximately in the middle of these ranges.The actual range of standard tightness, whencompare to its theoretical range, appears skewedtoward the direction of tight standards. The meanof job-related stress was in the middle of the the-oretical range. We had expected that these designengineers would have stressful work settings,which would imply that the mean would begreater than the middle of the theoretical range.

However, it could be that these engineers wereproviding relative, not absolute, assessments of their job-related stress. That is, if they had becomeaccustomed to high levels of job-related stress,they could have responded that they had ``aver-age'' job-related stress meaning that they hadexperienced a typical level of stress for the kind of job they have. The key, however, is not theirreported level of job-related stress but how it isassociated with the other variables. Table 2reports Pearson correlations for the variables. Thesigns of the correlations were consistent with sevenof the nine hypotheses ( p


11/18

(Fig. 1) and to test the hypotheses embedded ineach model. LISREL was used to simultaneouslyestimate a measurement model which related theobserved variables to the latent variables and astructural model which related the latent variables

to each other.13

In the measurement model,unobserved endogenous and exogenous latentvariables and their observed indicators were asso-ciated together through factor loadings and mea-surement errors. The relationship between anexogenous and an endogenous variable waslabeled as a g and the relationship between twoendogenous variables was labeled as b . The struc-tural error had an expected value of zero and wasuncorrelated with the exogenous variable. Themaximum-likelihood procedure was used to esti-mate all of the parameters and t indices. All of the variances for the observed indicators and

latent variables were estimated and all of the cov-ariances of these variables were xed at zero, withthe exception of some of the covariances among theerror terms for participative standard setting. 14

The LISREL model for the indirect model had a

satisfactory t to the data as evidenced by a com-parative t index (Bentler, 1990) of 0.90 and anincremental t index (Bollen, 1989) of 0.90. TheLISREL model for the direct model provided aslightly less-good t to the data as evidenced by acomparative t index of 0.88 and an incrementalt index of 0.88. Fig. 2 presents the structuralparameter estimates (i.e. g,b ) for the hypothesistests for each model. The factor loadings for eachvariable in each LISREL model had the expectedsign, reasonably large standard estimates, statis-tical signicance ( p


12/18

direct model was nested within the indirect model;all of its linkages involving job-related stress wereconstrained to be zero and the other linkages in

the two models also diered as identied in Fig. 2in terms of whether they were constrained to bezero or allowed to be estimated. The chi-squaredierence statistic between the direct and indirectmodels was 59.02 (d.f.=1, p


13/18

Overall, these test results indicate that standard-based incentives and standard tightness are inu-enced by the degree of subordinate participation

in standard setting, and that the eects of thesethree control-system components are indirect on job performance through job-related stress as theintervening variable. These results clarify andmodify the composite results implied by the prioraccounting studies. They reveal additional insightsthat can be obtained from researching indirecteects of control systems, especially using anindirect model such as the one in this studybecause it supports a dierent prediction for howto improve performance than does the directmodel. The dierence in prediction arises becausethe direct model predicts a positive relationshipbetween standard tightness and performance butthe indirect model predicts a negative relationship(i.e. predicted positive relationship between tight-ness and stress multiplied by the predicted negativerelationship between stress and performance). Themagnitude of these conicting predictions is heigh-tened in a sample such as the one in this studybecause job-related stress (the intervening variable)is a salient feature of the organizational context.

These results also help sort out the conicting

recommendations in the literature concerning howtight performance standards should be. The litera-ture has traditionally recommended that stan-dards/budgets should be ``tight, but achievable,''with a probability of less than 0.50 that they will beachieved [see Merchant and Manzoni (1989) for areview of this literature]. In contrast, Merchant andManzoni report eld-study evidence that protbudgets are achieved 8090% of the time. Con-sistent with Merchant and Manzoni's evidence, theindirect-eects model predicts that performance ishigher when performance standards are lower.While there may be practical limits to how low aperformance standard can be, our results indicatethat Merchant and Manzoni's evidence of, andrecommendation for, highly-achievable budgets isconsistent with the nding of a negative relation-ship between standard tightness and performanceoperating through job-related stress.

Merchant and Manzoni (1989) review the priorresearch which is used to support the recommenda-tion in the literature for high performance standards

and suggest that recommendation is valid when:(1) the subordinate is committed to achieving thestandard (which may be less likely if the standard

induces too much stress); (2) tasks have shortduration, low uncertainty and complexity, abilitydoes not explain most of the variance in perfor-mance, and there is immediate performance feed-back; and (3) performance is not aectedexogenous factors. Merchant and Manzoni thenconclude that since many work settings and per-formance (including that in our study) do not havethese features, the prior research which is used tosupport the recommendation of tight performancestandards does not provide a basis for any pre-scription. An implication is that future researchcould test whether these task and performancevariables as well as commitment aect the stan-dard tightness-performance relationship.

Limitations of this research study should beconsidered when interpreting its results. Theseinclude reliance on non-experimental data to testcausal relationships and on correlations which arenecessary but not sucient to establish causalrelations. The sample is a potential limitation interms of direct comparability to the samples usedin related research. However, since tests of predic-

tions that are based on those studies yielded similarresults, this should reduce concerns about general-izability or comparability. The measurement of thevariables, while possessing satisfactory measure-ment reliability, could be improved. 16 For example, job performance could be measured by reference toobjective measures (e.g. audited nancial-statementdata) or, if subjective measurement is relied on,then multiple raters of individual performancecould be used (e.g. superior, peers).

These results have implications for futureresearch beyond those already mentioned. Thedirect and indirect models were developed by pie-cing together the results of prior accounting stu-dies that individually typically investigated one ortwo of the nine hypotheses embedded in the directand indirect models. Future research could expandthe indirect model in order to develop a morecomplete understanding of the design and eectsof control systems. Expanded models can include,

16 See footnote 12.



14/18

for example, antecedents to control systems (e.g.uncertainty, information asymmetry, prior perfor-mance), additional control-system components

and their interrelationships (e.g. controllability of performance measures, audits, intra-period andend-of-period revisions of standards, formula vs.subjective performance evaluation), other inter-vening variables (e.g. goal acceptance, goal com-mitment, motivation 17 ), other dependent variables(e.g. innovation, risk taking, gaming behavior),and consequent variables (i.e. factors aected bydependent variables such as promotions, futureparticipation strategy). Future research also couldcontinue to use structural equation modelsbecause they facilitate testing, for example, com-plex nomological networks that would result fromincluding additional variables, as just suggested.Such models also provide simultaneous tests of measurement and structure and tests of competingmodels. Additional research that aims to expandand enrich our indirect model (Figs. 1 and 2) willprovide management accounting researchers witha better basis to understand the antecedents,design, and eects of control systems.

Acknowledgements

We are indebted to Chee Chow for his helpduring this project, and we appreciate the com-ments on earlier drafts of this paper by Jake Birn-berg, Harry Evans, Annie Farrell, Sue Haka,Vicky Homan, Dave Larcker, Joan Luft,Michael Maher, Nandu Nagarajan, Neale O'Con-nor, two anonymous reviewers, and the seminarparticipants at Arizona State University, BostonAccounting Research Colloquium, Georgia StateUniversity, Michigan State University, Universityof Pittsburgh, and Saint Louis University.

Appendix A. Description of the job/task

The sample in this study is automobile designengineers operating under a target costing strategy

in which a control system has performance goals,and their performance measures are stated innancial and nonnancial terms. The target cost-

ing system puts continuous pressure on the designengineers to improve their design process anddesigns. There is pressure to decrease the cost of and time to make designs and to increase thequality of their design process. There is also pres-sure to design automobile components that are of increasingly higher quality, functionality, and per-formance, and that have lower cost and less timeto market. We expect that this performance pres-sure causes job-related stress.

All design engineers work in Japan at the sitewhere this company does its automotive designengineering. The automobile division of this com-pany has a matrix organizational design consistingof functional (e.g. mechanical engineering, electricalengineering) departmental managers and productmanagers. The functional managers have moreimpact than the product managers on the designengineers' performance standards and performanceevaluation. Each engineer typically works on a par-ticular section of a car (e.g. chassis, engine). Thiscompany has had a long-term employment policyand design engineers spend most, if not all, of their

careers in automobile design engineering. Con-sequences of this organizational design includeengineers having a variety of job assignments withinthe design engineering departments and engineeringmanagers frequently having had one or more of the jobs which they currently manage. Consequences of this employment context are that the engineeringmanagers tend to be very familiar with their sub-ordinates' jobs and may know more about a parti-cular job than does a subordinate who is new to it.In this context, a participatively-oriented controlsystem is used to share information so that thesuperior can learn more about the subordinates' jobs and the subordinates can learn how to do their jobs better, and to motivate the design engineers toachieve their performance standards.

Subordinate engineers have frequent commu-nications with their functional and product engi-neering managers which are intended to identifyways to improve performance, to establish perfor-mance standards, and to monitor performancerelative to the standards. Performance standards

17 See footnote 7 for discussion of two studies that examinethe intervening eects of commitment and motivation.



15/18

(targets, goals) are set both for the design process,including design time (e.g. milestones) and designcost, and for the design object, including target

cost, functionality, styling, performance, quality(e.g. defect rates), manufacturability, and service-ability. Since these standards are tailored to eachdesign assignment and design object, a commonset of performance goals and measures does notexist. These standards typically are participativelyset and are applied to departments or teams of engineers. They are also an important part of theinformation set that is used to evaluate andreward individual design engineers. Although thesetting of performance standards involves severalparties (e.g. human resource management, topmanagement) besides the subordinate engineersand their functional and product managers, theseother parties have much less involvement with andinuence on standard setting. The target cost sys-tem results in the setting of performance standardsfor the design engineers that are tight and chal-lenging. Finally, while these design engineersreceive bonuses that are determined by severalfactors (e.g. team and division performance) andby several employees (e.g. superiors, peers, humanresource management), an increasingly important

determinant of these bonuses is individual perfor-mance relative to the performance standards.

Appendix B. Variable measurement

B.1. Participative standard setting

(Anchored by 1=Extremely low, and 7=Extre-mely high.)

1. The extent to which your superiors soughtyour input into the setting of your perfor-mance standards.

2. The extent to which your superiors soughtyour input into determining the amount of resources needed for your design assignments.

3. The importance that your superiors placedon including changes you had suggested inyour performance standards.

4. The importance that your superiors placedon including changes you had suggested into

the determination of the amount of resourcesto be provided for your design assignments.

5. The importance that your superiors placed

on not nalizing your performance standardsuntil you were satised with them.6. The importance that your superiors placed

on not nalizing the amount of resources tobe provided for your design assignmentsuntil you were satised with them.

7. Overall, the inuence that you had in settingyour performance standards.

8. Overall, the inuence that you had in deter-mining the amount of resources to be pro-vided for your design assignments.

B.2. Standard-based incentives


1. The degree to which valued rewards todesign engineers increase with increases intheir measured performance.

2. The degree to which design engineers' valuedrewards are totally determined by measuredperformance relative to performance standards.

3. Consider the design engineers whose perfor-mance relative to the performance standardsare in the top 25% of all design engineers'performance. The extent to which these engi-neers receive larger valued rewards than dothose engineers whose performance in rela-tion to the standards are not in the top 25%.

B.3. Standard tightness

1. Please indicate the total amounts of each of the following resources you believed typicallywere required for you to achieve your per-formance standards on design assignments;minus

2. Please indicate the total amounts of each of the following resources that typically weremade available to you on design assignments.

(Each anchored by 1= no resources, and 7=anextremely large quantity of resources.)

Four resources:



16/18

a. Your time at workb. Design technology (e.g. CAD, CAM, CAE)c. Assistance from other employees of your

company (e.g. time, skills, information)d. Assistance from external sources (e.g. con-sultants, suppliers, customers)

B.4. Job-related stress

For each statement, please circle the numberwhich most closely represents your true beliefsabout your work assignments.

(Anchored by 1=Never; 2=Rarely; 3=Some-times; 4=Often; 5=Almost always.)

1. Feeling that you have too little authority tocarry out the responsibilities assigned to you.

2. Being unclear on just what the scope andresponsibilities of your job are.

3. Feeling that you have too heavy a work load,one that you can't possibly nish during anordinary workday.

4. Thinking that you'll not be able to satisfy theconicting demands of various people overyou.

5. Not knowing what your supervisor thinks of

you, how he evaluates your performance.6. The fact that you can't get information nee-

ded to carry out your job.7. Feeling unable to inuence your immediate

supervisor's decisions and actions that aectyou.

8. Not knowing just what the people you workwith expect of you.

9. Thinking that the amount of work you haveto do may interfere with how well it gets done.

B.5. Job performance


1. The level of my measured performance rela-tive to my performance standards.

2. The level of my measured performance rela-tive to other design engineers' measured per-formance.

3. The level of my measured performance.

References

Anderson, J., & Gerbing, D. (1984). The eects of samplingerror on convergence, improper solutions and goodness-of-tindices for maximum likelihood conrmatory factor analysis.Psychometrika, 49 , 155173.

Atkinson, A. (1978). Standard setting in an agency. Manage-ment Science, 24 , 13511361.

Bagozzi, R. (1980). Causal models in marketing . New York:John Wiley and Sons, Inc.

Baiman, S. (1982). Agency research in managerial accounting:a survey. Journal of Accounting Literature, 1 , 154.

Baiman, S., & Evans, J. (1983). Pre-decision information andparticipative management control systems. Journal of Accounting Research, 21 , 371395.

Beehr, T. (1985). Organizational stress and employee eective-ness. In Beehr, T. & Bhagat, R. Human stress and cognition inorganizations . New York: John Wiley and Sons.

Beehr, T., & Bhagat, R., (1985). Introduction to human stress andcognitionin organizations.InBeehr,T. & Bhagat,R. Humanstressandcognition in organizations . New York: JohnWiley and Sons.

Bentler, P. (1990). Comparative t indexes in structural models.Psychological Bulletin, 107 , 238246.

Birnberg, J., Shields, M., & Young, S. M. (1990). The case formultiple methods in empirical management accountingresearch (with an illustration from budget setting). Journal of Management Accounting Research, 2 , 3366.

Bollen, K. (1989). Structural equations with latent variables .New York: John Wiley and Sons.

Bommer, W., Johnson, J., Rich, G., Podsako, P., & Mack-enzie, S. (1995). On the interchangability of objective and

subjective measures of employee performance: a meta-analy-sis. Personnel Psychology, 48 , 587605.

Brownell, P., & Hirst, M. (1986). Reliance on accountinginformation, budgetary participation, and task uncertainty:tests of a three-way interaction. Journal of AccountingResearch, 24 , 241249.

Brownell, P., & McInnes, M. (1986). Budgetary participation,motivation, and managerial performance. The AccountingReview, 61 , 587600.

Chow, C. (1983). The eects of job standard tightness andcompensation scheme on performance: an exploration of linkages. The Accounting Review, 58 , 667685.

Chow, C., Cooper, J., & Waller, W. (1988). Participative bud-geting: eects of a truth-inducing pay scheme and informa-tion asymmetry on slack and performance. The AccountingReview, 63 , 111122.

Chow, C., Cooper, J., & Haddad, K. (1991). The eects of payschemes and ratchets on budgetary slack and performance: amultiperiod experiment. Accounting, Organizations and Society, 16 , 4760.

Christensen, J. (1982). The determination of performance stan-dards and participation. Journal of Accounting Research, 20 ,589603.

Collins, F. (1978). The interaction of budget characteristics andpersonality variables with budgetary response attitudes. TheAccounting Review, 53 , 324335.



17/18

Cyert, R., & March, J. (1963). A behavioral theory of the rm .Englewood Clis, NJ: Prentice Hall.

Demski, J., & Feltham, G. (1978). Economic incentives in bud-getary control systems. The Accounting Review, 53 , 336359.

Dillard, J., & Fisher, J. (1990). Compensation schemes, skilllevel and task performance: an experimental investigation.Decision Sciences, 21 , 121137.

Dunk, A. (1993). The eects of job-related tension on manage-rial performance in participative budgetary settings.Accounting, Organizations and Society, 18 , 575585.

Dunk, A., & Nouri, H. (1998). Antecedents of budgetary slack:a literature review and synthesis. Journal of Accounting Lit-erature, 17 , 7296.

Edwards, J. (1996). An examination of competing versions of the person-environment t approaches to stress. Academy of Management Journal, 39 , 292339.

Furnham, A., & Stringeld, P. (1994). Congruence of self andsubordinate ratings of managerial practices as a correlate of superior evaluation. Journal of Occupational and Organiza-tional Psychology, 67 , 5767.

Govindarajan, V. (1986). Impact of participation in the bud-getary process on managerial attitudes and performance:universalistic and contingency perspectives. Decision Sci-ences, 17 , 496516.

Harrison, G. (1992). The cross-cultural generalizability of therelation between participation, budget emphasis and job-related attitudes. Accounting, Organizations and Society, 17 ,115.

Heneman, H. (1974). Comparisons of self- and superior-ratingof managerial performance. Journal of Applied Psychology,59, 638642.

Hertenstein, J., & Platt, M. (1998). A management accountingchallenge: designing performance measures for industrialdesign and new product development teams. Working paper,Northeastern University.

Hirst, M., & Lowy, S. (1990). The linear additive and inter-active eects of budgetary goal di culty and feedback onperformance. Accounting, Organizations and Society, 15 ,425436.

Hofstede, G. (1967). The game of budget control. Assen . TheNetherlands: Van Gorcum.

Hopwood, A. (1976). Accounting and human behavior . Engle-wood Clis, NJ: Prentice Hall.

Horngren, C., Foster, G., & Datar, S. (1997). Cost accounting:a managerial emphasis (9th ed.). Upper Saddle River: Pren-

tice Hall.Jamal, M. (1985). Relationships of job stress to job perfor-

mance: a study of managers and blue-collar workers. HumanRelations, 38 , 409424.

Jex, S., & Beehr, T. (1991). Emerging theoretical and metho-dological issues in the study of work-related stress. Researchin Personnel and Human Resources Management, 9 , 311365.

Jick, T. (1984). The stressful eects of budget cuts in organiza-tions. In Rosen, L. Topics in managerial accounting (3rd ed.),Toronto: McGrawHill Ryerson.

Jick, T. (1985). As the ax falls: budget cuts and the experienceof stress in organizations. In Beehr, T. & Bhagat, R. Human

stress and cognition in organizations . New York: John Wileyand Sons.

Joreskog, K., & Sorbom, D. (1979). Advances in factor analysisand structural equation models . Cambridge: MA: ABT.

Kahn, R., Wolfe, D., Quinn, R., Snoek, J., & Rosenthal, R.(1964). Organizational stress: studies in role con ict and ambiguity . New York: John Wiley.

Kenis, I. (1979). Eects of budgetary goal characteristics onmanagement attitudes and performance. The AccountingReview, 54 , 707721.

Kirchner, W. (1965). Relationships between supervisory andsubordinate ratings for technical personnel. Journal of Industrial Psychology, 3 , 5760.

Kren, L. (1990). Performance in a budget-based control system:an extended expectancy theory model approach. Journal of Management Accounting Research, 2 , 100112.

Kren, L., & Liao, W. (1988). The role of accounting informa-tion in the control of organizations: a review of the evidence.Journal of Accounting Literature, 7 , 280309.

Lau, C., Low, L., & Eggleton, I. (1995). The impact of relianceon accounting performance measures on job-related tensionand managerial performance: additional evidence. Account-ing, Organizations and Society, 20 , 359381.

Lindquist, T. (1995). Fairness as an antecedent to participativebudgeting: examining the eects of distributive justice, pro-cedural justice and referent cognitions on satisfaction andperformance. Journal of Management Accounting Research,7 , 122147.

Locke, E., & Latham, G. (1990). A theory of goal setting and task performance . Englewood Clis NJ: Prentice-Hall.

Lowe, E., & Shaw, R. (1968). An analysis of managerial bias-

ing: evidence from a company's budgeting process. TheJournal of Management Studies, 5 , 304315.Lukka, K. (1988). Budgetary biasing in organizations: theore-

tical framework and empirical evidence. Accounting, Organi-zations and Society, 13 , 281301.

Mahoney, T., Jerdee, T., & Carroll, S. (1965). The job(s) of management. Industrial Relations, 4 , 97110.

McGrath, J. (1976). Stress and behavior in organizations. InM. Dunnette Handbook of industrial and organizational psy-chology . Chicago: Rand McNally.

Merchant, K. (1981). The design of the corporate budgetingsystem: inuences on managerial behavior and performance.The Accounting Review, 56 , 813829.

Merchant, K. (1982). The control function of management.

Sloan Management Review, 23 , 4355.Merchant, K. (1985). Budgeting and the propensity to create

budgetary slack. Accounting, Organizations and Society, 10 ,201210.

Merchant, K. (1989). Rewarding results: Motivating prot cen-ter managers . Boston: Harvard Business School.

Merchant, K., & Manzoni, J. F. (1989). The achievability of budget targets in prot centers: a eld study. The AccountingReview, 64 , 539558.

Milani, K. (1975). Budget-setting, performance and attitudes.The Accounting Review, 50 , 274284.

Motowidlo, S., Pacard, J., & Manning, M. (1986). Occupational



18/18

stress: its causes and consequences. Journal of Applied Psy-chology, 71 , 618629.

Nixon, B. (1998). Research and development performance mea-surement: a case study. Management Accounting Research, 9 ,

329355.Nouri, H., & Parker, R. (1998). The relationship between bud-

get participation and job performance: the roles of budgetadequacy and organizational commitment. Accounting,Organizations and Society, 23 , 467483.

Onsi, M. (1973). Factor analysis of behavioral variablesaecting budgetary slack. The Accounting Review, 48 , 535 548.

Parker, J., Taylor, E., Barrett, R., & Martens, L. (1959). Ratingscale content: III. relationship between supervisory and self-ratings. Personnel Psychology, 12 , 4963.

Riggio, R., & Cole, E. (1992). Agreement between subordinateand superior ratings of supervisory performance and eectson self and subordinate job satisfaction. Journal of Occupa-tional and Organizational Psychology, 65 , 151158.

Rockness, H. (1977). Expectancy theory in a budgetary setting:an experimental examination. The Accounting Review, 52 ,893903.

Ronen, J., & Livingstone, J. (1975). An expectancy theoryapproach to the motivational impacts of budgets. TheAccounting Review, 50 , 671685.

Schi, M., & Lewin, A. (1970). The impact of people on bud-gets. The Accounting Review, 45 , 259268.

Shields, J., & Shields, M. (1998). Antecedents of participativebudgeting. Accounting, Organizations and Society, 23 , 4976.

Shields, M., & Waller, W. (1988). A behavioral study of accounting variables in performance-incentive contracts.

Accounting, Organizations and Society, 13 , 581594.Shields, M., & Young, S. M. (1993). Antecedents and con-sequences of participative budgeting: evidence on the eectsof asymmetrical information. Journal of ManagementAccounting Research, 5 , 265280.

Simons, R. (1988). Analysis of the organizational character-istics related to tight budget goals. Contemporary AccountingResearch, 4 , 267283.

Stedry, A. (1960). Budget control and cost behavior . Englewood

Clis, NJ: Prentice Hall.Van Harrison, R. (1985). The personenvironment t model

and the study of job stress. In Beehr, T. & Bhagat, R. Humanstress and cognition in organizations . New York: John Wileyand Sons.

Venkatraman, N., & Ramanujam, V. (1987). Measurement of business economic performance: an examination of methodconvergence. Journal of Management, 13 , 109122.

Waller, W. (1988). Slack in participative budgeting: the jointeect of a truth-inducing pay scheme and risk preferences.Accounting, Organizations and Society, 13 , 8798.

Waller, W., & Chow, C. (1985). The self-selection and eorteects of standard-based employment contracts: a frame-work and some empirical evidence. The Accounting Review,60, 458476.

Weick, K. (1983). Stress in accounting systems. The AccountingReview, 58 , 350369.

Williamson, O. E. (1964). The economics of discretionary beha-viour: managerial objectives in a theory of the rm . Engle-wood Clis, NJ: Prentice Hall.

Young, S. M. (1985). Participative budgeting: the eects of riskaversion and asymmetric information. Journal of AccountingResearch, 23 , 829842.

Young, S. M. (1988). Individual behavior: performance, moti-vation, and control. In Ferris, K. Behavioral accountingresearch: A critical analysis . Columbus, OH: Century VIIPublishing Company.

Young, S. M., & Lewis, B. (1995). Experimental incentive con-tracting research in management accounting. In Ashton, R.& Ashton, A. Judgment and decision making research inaccounting and auditing . Cambridge, UK: Cambridge Uni-versity Press.


shields deng kato2000

Documents