Performance 3

An Experimental Investigation of Employer Discretion in Employee Performance Evaluation and
Compensation
Author(s): Joseph G. Fisher, Laureen A. Maines, Sean A. Peffer and Geoffrey B. Sprinkle
Source: The Accounting Review, Vol. 80, No. 2 (Apr., 2005), pp. 563-583
Published by: American Accounting Association
Stable URL: http://www.jstor.org/stable/4093069
Accessed: 07-03-2015 10:43 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
American Accounting Association is collaborating with JSTOR to digitize, preserve and extend access to The Accounting
Review.
http://www.jstor.org
This content downloaded from 121.52.157.189 on Sat, 07 Mar 2015 10:43:31 UTC
All use subject to JSTOR Terms and Conditions
THE ACCOUNTING REVIEW
Vol. 80, No. 2
2005
pp. 563-583
An
ExperimentalInvestigation
of Employer Discretion in
Employee Performance Evaluation
and Compensation
Joseph G. Fisher
Indiana University
Laureen A. Maines
Indiana University
Sean A. Peffer
Universityof Kentucky
Geoffrey B. Sprinkle
Indiana University
ABSTRACT:Employmentrelationshipsprovide fertileground for both employee and
employer opportunism.Employersworry about whether employees will devote suffi-
cient effortto work,and employees are concerned about whetheremployerswillcom-
pensate them appropriately.In this paper, we examine whether employer discretion
over the size of the total employee compensation pool and the allocationof this pool
among employees influences employee and employer opportunism.The results of our
experiment indicate that firmoutput and employees' compensation are greaterwhen
the employer does not have discretion over total employee compensation, but does
have discretion over the allocationof total compensation. We find that the employer's
residual profit increases with discretion over the allocation of compensation among
employees; however, we find no effect on residualprofitof the employer'sdiscretion
over the total amountof employee compensation. Ourresultssuggest that firmsbenefit
from a compensation contractthat establishes total employee compensation as a pre-
determined function of public, aggregate measures such as accounting income, but
We thankWaltBlacconiere,Dave Barrett,Daniel Beneish,Bob Bowen, Dave Greene,JohnHassell,Dan Heitger,

FrankHodge,JaneJollineauKennedy,JamiePratt,JerrySalamon,TerryShevlin,ReedSmith,MichaelWilliamson,
two anonymousreviewers,and participantsat researchworkshopsat EmoryUniversity,IndianaUniversityBloom-
ington, and IndianaUniversityIndianapolis,Universityof Maryland,MichiganState University,Universityof
Missouri,The Universityof NorthCarolinaat ChapelHill, Universityof SouthCarolina,and Universityof Wash-
ingtonfor helpfulcommentsand suggestions.The Universityof Kentuckyand the PresslerFacultyFellowshipat
IndianaUniversityprovidedgenerousfundingfor this project.
Editor'snote:This paperwas acceptedby MarlysGaschoLipe, Editor.
Submitted July 2003
Accepted Sept 2004
563
564 Fisher Maines, Peffer,and Sprinkle
providesthe employerat least some discretionto allocatethis compensationusing

However,ourresultscautionthatemployeesand employersmay
privateinformation.
not have similarpreferences for the degree of employer discretion over the determi-
nation of total employee compensation.
Keywords: compensation;discretion;subjectiveperformanceevaluation;double-sided
moralhazard.
Data Availability:Contact the authors.
I. INTRODUCTION
mploymentrelationshipstypicallyare characterizedby two-sidedconcernsregarding
opportunisticbehavior(Demski 1997, 579). Employersare concernedaboutwhether
employees will work diligently, while employees worry about whether employers
will rewardthis work appropriately.The potentialfor employees and employers to act in
their own self-interest when making these choices threatensto destroy the cooperation
needed for productivefirm outcomes.
Opportunismmay not arise if workplacenormsresultin cooperationamong employers
and employees. However, firms typically do not rely solely on workplacenorms to pre-
empt opportunism,and implementcontrol systems, such as performanceevaluationand
compensationsystems, to mitigate opportunism.These systems govern two compensation
decisions: (1) how to allocate total remunerationbetween the employerand employees and
(2) how to allocate total employee remunerationamong employees (Baiman and Rajan
1995). The first decision establishesthe size of the total employee compensationpool and
the second determinesthe allocation of this pool among employees. The approacha firm
uses to make these decisions can influenceboth employerand employee opportunism.
In this paper,we examine whetheran employer'sdiscretionover the two compensation
decisions affects employee and employer opportunism.The effect of employer discretion
over the two compensationdecisions is unclear because discretion creates tension with
respect to employer and employee opportunism.An employer can use discretion to her
personalbenefit, increasingher share of total compensationat the expense of employees
or allocating compensationto "favored"employees.' However,discretionalso allows an
employer to reduce employee opportunismby determiningemployee compensationusing
measuresand approachesthat best reflect employees' collective and individualefforts.
The tension between employer discretion and employer/employeeopportunismtypi-
cally is connectedto firms' informationsystems. Firms'formal informationsystems, such
as accounting systems, provide informationthat is publicly observableor verifiablebut
often reflects only aggregateemployee performance(Baimanand Rajan 1995). Employers
can alleviateemployees'concernsaboutemployeropportunismby giving up discretionand
committingex ante to a total compensationpool and its allocationamongemployees based
on such public information.However,public informationmay not reflectall relevantaspects
of performance.In particular,aggregateinformationprovides little basis for employers to
differentiallyrewardindividualemployee performanceand preventemployees from free-
ridingon the effortsof otheremployees. Thus, compensationapproachesbased on publicly
available informationmay reduce employer opportunismbut may not prevent employee
opportunism.
Throughoutthe paper,we use feminine pronounsto refer to employersand masculinepronounsto refer to

employees.
The AccountingReview,April 2005
An ExperimentalInvestigationof EmployerDiscretion 565
Firms also have informalsources of information.This informationcan reflect multiple

aspects of aggregateperformanceand/or isolate the performanceof individualemployees,
but is privatelyobservedby the employer.An employercan reduceemployees'opportunism
by retainingdiscretionto use her privateinformationin performanceevaluation.For ex-
ample, an employercan reduce the benefit of free-ridingby using her privateinformation
on individualperformanceto allocate the compensationpool among employees. Unfortu-
nately, allowing the employer discretion to use private informationgenerates employee
concerns that the employerwill fail to rewardemployees appropriatelyfor their aggregate
or individualefforts.Thus, a compensationscheme in which an employerhas full discretion
to use private informationmay reduce employee opportunismbut allows for employer
opportunism.
The relativestrengthsand weaknessesof restrictingversusallowingemployerdiscretion
suggests that there may be complementaryroles for discretionversus nondiscretionwith
respectto the two compensationdecisions. Specifically,an approachin which the employer
does not have discretionover the size of the total employee compensationpool (hereafter
pool size) but does have discretionover the allocationof the employee compensationpool
(hereafterpool allocation)may alleviate concerns about both employerand employee op-
portunism(Baiman and Rajan 1995). This approachsuggests that there also are comple-
mentaryroles for public and privateinformation-public informationprovidesthe basis for
the total compensationpool and privateinformationassists in the allocationof this pool.
To investigatethe effects of employerdiscretionon employee and employeropportun-
ism, we examine situationsin which the employereitherhas full discretionor no discretion
over the size of the employee compensationpool and/or the allocation of this pool among
employees. We use a laboratoryexperimentwith a 2 (pool size discretion) x 2 (pool
allocation discretion) x 8 (period) mixed design. In the experiment,the employer either
has discretionto determinethe size of the two-employeecompensationpool (discretionary
pool size, DS) or the pool is set at two-thirdsof the revenuefrom total employee output
(nondiscretionarypool size, NDS). Similarly,the employereither has discretionin allocat-
ing the compensationpool between the two employees (discretionarypool allocation,DA)
or the pool is split evenly between the two employees (nondiscretionarypool allocation,
NDA). The employerreceives the residualprofitafter subtractingemployee compensation
from total revenue.
In the experiment,237 undergraduate businessstudents,in groupsof three,acted either
as an employeror as one of two employees. Employeesallocatedthree minutesbetween a
worktask (decodingnumbersinto letters)andcompensatedleisurefor each of eight periods.
After each work session, employersin the discretionaryconditionsdeterminedthe size of
the employee compensationpool and/or the allocationof this pool between the two em-
ployees. Employersreceivednoisy privateinformationaboutindividualemployee outputto
use if desiredin makingtheircompensationdecisions.Thus, in the discretionaryconditions,
employers had the opportunityto more closely align an employee's compensationwith
individualproductionthan in the nondiscretionaryconditions in which compensationwas
a multipleof total outputand/or evenly split between the two employees.
We consider the impact of employerdiscretionover pool size and pool allocationon
total output, as well as the division of the value of total output (i.e., revenue) between
employer remunerationand employee remuneration.The results indicate that both total
outputand employee compensationbenefit from restrictingthe employer'sdiscretionover
the size of the compensationpool, and allowing the employerto have discretionover the
allocation of this pool. Total output and employee compensationwere greater when the
TheAccountingReview,April 2005
566 Fisher, Maines, and Sprinkle
Peffer,
employerdid not have discretionover pool size than when she did have discretion,as well
when the employerhad discretionover the pool allocationthan when employees received
equal compensation.These results hold even when we eliminate the last two periods to
control for possible end-gameeffects.
Ourresultsfor the effect of discretionon the employer'sresidualprofitare mixed.Over
the entire eight periods, the employer's residualprofit did not differ between the discre-
tionary and nondiscretionarypool size conditions. The employer's residual profit was
greaterin the discretionarypool allocationconditionthanin the nondiscretionary condition.
However,we found no effect of eitherpool size or pool allocationdiscretionon employer's
residualprofitaftereliminatingthe final two periods.This resultsuggests thatdiscretionary
pool allocation primarilymitigates employee opportunismthat arises as the employment
relationshipnears its end.
Our results also documenta fairly high level of cooperativebehaviorbetween employ-
ers and employees in the absence of controls that mitigate opportunism.In the DS-NDA
contract,which provides no explicit controls to mitigateemployee or employer opportun-
ism, employees produced,on average, 75 percent of the amount produced in the best-
performingNDS-DA contract.Additionally,employerswith discretionover pool size paid
out, on average,59 percentof revenuesas compensationto the two employees. However,
the effects of opportunismassociated with either a discretionarypool size or nondiscre-
tionarypool allocationapproachincreasedover the eight periods.For example, in the last
period,outputin the DS-NDA contractwas only 45 percentof thatin the NDS-DA contract.
Moreover,employersin the two discretionarysize conditionspaid only 51 percentof rev-
enue to employees in the last period.
This study contributesto the control systems literatureby examining implicationsof
employer discretion in performanceevaluationand compensationfor employer and em-
ployee opportunism.Our results documenta fairly high level of productiveoutputin the
absence of specific controls relatedto employerdiscretion/nondiscretionintendedto miti-
gate employee and employer opportunism.Rather,employers and employees appear to
recognizethe economic benefitsof joint cooperationand rely on social values, such as trust
and reciprocity,to keep productiveoutcomes intact over all but the final periods of em-
ployment (Axelrod 1984; Berg et al. 1995; Sprinkle2003). These findings highlight that
the efficacy and cost of performanceevaluationand compensationarrangementsshould be
compared not simply against each other, but against the cost and efficacy of naturally
occurringsocial mechanismsin the workplace.
Despite these cooperativeoutcomes, our study documentsbenefits of control systems
with explicit incentives to mitigate employer and employee opportunism.Specifically,al-
lowing the employerdiscretionto allocate compensationamong employees increasesfirm
output and both employee and employer remuneration.However,the implicationsof our
resultsfor discretionin settingthe size of total employeecompensationare less clear.While
eliminatingemployerdiscretionover the pool size improvesoutputand employee compen-
sation, the employer's residual profit is not greaterunder the nondiscretionarypool size
case than underthe discretionarypool size case. In essence, the benefitto the employerof
increasedoutput associated with a nondiscretionarypool size is not sufficiently large to
offset the cost to the employerof havingto pay greaterper unit compensationto employees
in our experiment.Thus, employersmay preferto retaindiscretionover both pool size and
pool allocation,to the detrimentof firmoutputandemployeecompensation.However,given
the compensation-related benefitsto employees of a nondiscretionaryapproachto pool size,
competitionamong employersfor employees could lead employersto relinquishtheir dis-
cretion over compensationpool size.
The Accounting Review, April 2005
Finally, our study complements research on the use of contractible(public) versus

noncontractible(private)informationin compensationcontracts.Analyticalresearchdocu-
ments the benefits of compensationcontractsthat base the total pool size on contractible
informationand the pool allocationon employer'sprivateinformation(Baimanand Rajan
1995). Our results generally provide empirical support for Baiman and Rajan's (1995)
conclusion.Since contractibleinformationoften emanatesfrom formalaccountingsystems,
our findings also suggest that there are complementaryroles in designing compensation
systems for informationfrom formal accountingsystems and privateinformationfrom a
firm's informalinformationsystems.
The remainderof this paperis organizedinto four sections. The next section provides
backgroundand develops the hypotheses,and Section III describesthe experimentused to
test these hypotheses.Section IV presentsthe results, and Section V providesa summary
and discussion of the results.
II. BACKGROUNDAND HYPOTHESES

Employer and Employee Opportunism-Discretionary Pool Size/Nondiscretionary
Pool Allocation Contract (DS-NDA)
We examine a setting in which employees work independentlyon a productiontask
for a finite numberof periods.Each employee allocatesa fixed amountof time each period
between work and leisure, and derives utility from both leisure time and compensation.
Outputfor the groupof employees is the sum of individualemployee output.Firmrevenue
equals groupoutputmultipliedby the competitivemarketprice. The employeris a residual
claimant who receives total firm revenue less the amount paid as compensation to
employees.
The firm has an informationsystem that measuresemployee performanceand a com-
pensation system that rewardsemployee performance(Zimmerman2003, 169). To study
our researchquestionon the effects of employerdiscretionon opportunism,we vary char-
acteristicsof these two systems. We firstexaminea settingthatprovidesno formalcontrols
to mitigateeitheremployeror employee opportunismand then add controlsthatpotentially
mitigateone or both types of opportunism.
In our initial setting, the informationsystem measuresgroup output (e.g., revenue or
income for the entire company or a division), but not individualemployee output.2The
employerretainsdiscretionto determinethe size of employees' compensationpool. Given
the lack of informationon individualperformance,the employerdoes not retaindiscretion
over the allocation of the compensationpool. Rather,the pool is divided equally among
the employees. We term this compensationsystem the discretionarysize/nondiscretionary
allocation(DS-NDA) contract.
These choices for informationand compensationsystems representcompany settings
that providefertile groundfor both employerand employee opportunism.Employeeshave
an incentiveto act opportunisticallyand shirk in their work effort because their individual
compensationis based on group performance.If an employee receives more utility from
leisure than from the expected incremental compensation he would receive from working,
2
In practice, employers often observe aggregate performance measures from firms' formal information systems,
but can not disentangle individual employee performance due to complexities of the production process. In these
settings, employers frequently use aggregated performance as a basis for incentive contracts and allocate total
compensation in a manner unrelated to individual performance (Milgrom and Roberts 1992, 413).
568 Maines,Peffer and Sprinkle
Fisher,
then he may shirk and receive work compensationbased on the efforts of his co-workers
(free-ride).3
Employersalso have an incentive to act opportunisticallygiven their discretionover
the size of total employee compensation.Given a finite-periodworld, a self-interestedem-
ployer will pay no employee compensationin the last period since it only reduces her
residualpaymentand has no incentive value for employees' futurework-leisureallocation
decisions (employees have alreadymade all such decisions). If employees believe the em-
ployer will not pay compensationin the final period, then they have no incentiveto work
and will allocate all time to leisure. Backwardinductionimplies a Nash equilibriumin
which the employerdoes not pay compensationand employees do not work in any period.4
Both employerand employee opportunismthreatento destroyproductiveoutcomesthat
could benefit all firm members.This threatarises from employerdiscretionover total em-
ployee compensationand employee discretionover work effort, combined with incentives
for employersand employees to use their discretionfor personalgain.
The threatof opportunismmay not materializeif employees and employerschoose to
cooperatewith one another,either out of recognitionof their mutualbenefitor out of fear
that others will retaliate against opportunisticbehavior (Fehr and Gichter 2000). Such
cooperativebehaviorrequirestrust that others will not act opportunistically.Researchon
trust games shows that individualsare willing to place themselves at risk by cooperating
and trustingthatotherindividualswill not exhibitself-interestedbehavior(Berg et al. 1995).
If all employees initiallytrustthatotheremployees will choose to workratherthanto shirk,
then a productiveequilibriumcan be sustainedas long as all employees reciprocate.Indeed,
Fisheret al. (2003) find thatemployees' work allocationsand performanceare significantly
greaterthan zero in a setting that does not provide a compensation-basedmechanismto
mitigateemployee opportunism.5
Employees also must trustthat the employer will not act in her own self-interest,but
instead will reciprocateby compensatingemployees appropriately.This expectationis not
withoutfoundation,as researchindicatesthat employersrewardemployees who exert high
levels of effort (Fehr et al. 1997). Additionally,over time, an employer can develop a
reputationfor compensatingemployees that induces employees to work in advanceof re-
ceiving compensation(Bull 1987).
Thus, it is possible that extreme opportunismwill not arise in the DS-NDA contract
due to social values such as trust and reciprocity.However,if these social values are not
sustainableworkplacenorms,then other mechanismsare needed to mitigateemployee and
employer opportunism.The firm's performanceevaluationand compensationsystem can
serve as such a mechanism.In the next section, we use implicationsof a paperby Baiman
and Rajan(1995) to considerhow changes to the employer'sdiscretionin settingemployee
compensationcan help to mitigate employer and employee opportunism.Specifically,we
examine three compensationcontractsthat change one or both aspects of the DS-NDA
SThis is similarto publicgoods settingsin which individualshave the incentivenot to contributeto the common
pool, but consumethe benefitsof contributionsfromotherindividuals(Ledyard1995).
4 Forexample,the employerknowsthatpayingcompensation in the next-to-lastperiodcannotmotivateemployees'
effort for that period since productionhas alreadyoccurred,but could motivateemployees'effort in the last
period.However,the employerknows that employees will not work in the last periodgiven that she can not
committo payingcompensation.Thus, thereis no incentivefor her to pay compensationin the next-to-the-last
period.
SHowever,Fisheret al. (2003) includea productiongoal in theirsetting,which may act as a separatemechanism
influencingemployeebehavior.
An Experimental Investigation of Employer Discretion 569
contractwith respect to employer discretion/nondiscretionin determiningthe size of the

employee compensationpool and/or the allocationof this pool among employees.
Compensation Contracts to Mitigate Employer and Employee Opportunism

Baiman and Rajan(1995) provide insights on featuresof compensationcontractsthat
have potentialto mitigateemployee and employeropportunism.Baimanand Rajan(1995).
(hereafterB-R) examine a contractin which the employer commits ex ante to determine
the size of the employee compensationpool based on jointly observable (contractible)
information,but does not commit to a specific pool allocation. Rather,the employer has
access to private(noncontractible)informationaboutan individualemployee'sperformance
and retainsthe right to use this informationto allocate the compensationpool. B-R ana-
lytically show that this contractresultsin a strictParetoimprovementover a compensation
contractin which the employer commits to determiningcompensationusing only jointly
observableinformation.
The findingsin B-R highlightthe importanceof employerdiscretionfor employee and
employeropportunismassociatedwith the DS-NDA contract.The optimalcontractin B-R
has two features that differ from the DS-NDA contract. First, in the B-R contract,the
employer gives up discretionto set the compensationpool size ex post. Second, the em-
ployer retains discretion to allocate the compensationpool among employees using her
private informationrelated to individual employee performance.Thus, B-R's analytical
resultssuggest thatemployeeandemployeropportunismcan be mitigatedby choices related
to employer discretion. We discuss these issues below for employee and employer
opportunism.
Mitigating Employee Opportunism-Discretionary Pool Size/DiscretionaryPool

Allocation Contract(DS-DA)
In the DS-DA contract,the employerhas complete discretionover both the amountof
the compensationpool and the allocationof the pool among employees. While the DS-DA
contractdoes not alleviate employer opportunismdue to this discretion,the contractpro-
vides an opportunityfor mitigatingemployee opportunism.Specifically,if the information
system is expandedto include the employer'sprivatenoisy signals about each employee's
output (as in B-R), the employer can use her discretion to determinethe allocation of
compensationamong employees based on privateinformationabout individualemployee
performance.6By more closely aligning an employee's compensationwith his individual
performance,the employerreducesthe benefitsof shirkingto the employee, which should
increasehis work efforts.
Mitigating EmployerOpportunism-NondiscretionaryPool Size/NondiscretionaryPool

Allocation Contract(NDS-NDA)
In this contract,the employer has no discretionin setting either the size of the com-
pensationpool or its allocationamong employees; in essence, the employerhas no role in
determiningemployee compensation.The total employee compensationpool is a function
of output(percentageof total revenue)and employees receive an equal share of the com-
pensationpool. The equal division of the compensationpool among employees provides
6 The characterization of the employer'sprivateinformationas a noisy signalof individualemployeeperformance
reflectsthe idea that individualemployeeoutputtypicallyis not easily identified(Prendergast1999, 11). How-
ever, employersoften receive informationrelatedto aspects of individualemployee performance.This infor-
mationtypicallyis privateand based on subjectiveassessments.If the signals were publicratherthanprivate,
then an employercould explicitlytie an employee'scompensationto his individualperformancesignal.
570 Maines, Peffer,and Sprinkle
Fisher,
no formalcontrolsfor employee opportunism;however,the NDS-NDA contractdoes mit-

igate employeropportunismby taking away the employer'sdiscretion.Note that the NDS-
NDA contractis more extremein this respectthanthe optimalB-R contract.The employer
has no discretionin the NDS-NDA contract,while the employerin the B-R contractretains
discretion over the parametersof the compensationpool but commits ex ante to these
parametersusing jointly observableinformation.Thus, the NDS-NDA contractfocuses on
eliminating employer opportunism,as opposed to creating an optimal contractfrom the
employer'sviewpoint,as in B-R.7
Mitigating Both Employeeand Employer Opportunism-NondiscretionaryPool Size/

DiscretionaryPool Allocation Contract(NDS-DA)
In the NDS-DA contract,the compensationpool is a predeterminedfunctionof group
output,but the employerhas discretionin allocatingthe pool among employees. Employer
opportunismis mitigatedbecause the employerhas no discretionover the size of the com-
pensationpool. Moreover,the employer can reduce employee opportunismby using her
privatesignals aboutindividualperformanceto align an employee's compensationwith his
performance.Thus, the NDS-DA contract has potential to mitigate both employer and
employee opportunism.
Hypotheses
The priordiscussion indicatesthat use of a nondiscretionaryapproachto determining
the size of the employee compensationpool can mitigateemployeropportunism.This in-
creases employees' willingness to devote time to work, which, in turn, increases output.
Similarly,the use of a discretionaryapproachto allocatingthe employeecompensationpool
should mitigateemployee shirkingand increaseoutput.Employees will devote more time
to work because they expect that their employerwill exercise her discretionand use infor-
mation about their individualproductionto determinetheir share of total employee com-
pensation.Ceterisparibus,bothemployees andemployersshouldbenefitfromthe increased
output.These argumentslead to the following hypotheses.
Employees'GroupOutput(Performance)8
Hla: Group output will be greaterwhen the employer does not have discretionover
the size of the employee compensationpool than when the employerhas discre-
tion over the size of the employee compensationpool.
Hlb: Groupoutputwill be greaterwhen the employerhas discretionover the allocation

of the employee compensationpool than when the employerdoes not have dis-
cretion over the allocationof the employee compensationpool.
7 Thereare severalotherdifferencesbetweenB-R and our paper.First,the B-R model is a single-periodmodel

while our setting includes a finitely repeatedsetting. Second, in the B-R model, outcomes and signals are
dichotomous,while in our settingproductionoutcomesand signalscan takeon numerouslevels. Finally,in the
B-R model, therearejointly observablesignals abouteach employee'sindividualperformance,whereasin our
settingall individualsignalsare the employer'sprivateinformationandonly groupoutputis jointly observable.
Thus, while B-R has implicationsfor our setting,our paperis not a directtest of theiranalyticalpredictions.
X Predictionssimilar to Hla and Hlb could be made for employees' allocationof time to work. Given our
assumptionthat outputis a functionof work time, we providehypothesesonly for output(performance).We
reportempiricalresultsfor employees'work time in footnote 13.
Employees' Compensation
H2a: Employees'compensationwill be greaterwhen the employerdoes not have dis-

cretionover the size of the employee compensationpool than when the employer
has discretionover the size of the employee compensationpool.
H2b: Employees'compensationwill be greaterwhen the employerhas discretionover
the allocationof the employee compensationpool than when the employerdoes
not have discretionover the allocationof the employee compensationpool.
Employer's Residual Profit
H3a: The employer'sresidualprofitwill be greaterwhen the employerdoes not have

discretionover the size of the employee compensationpool than when the em-
ployer has discretionover the size of the employee compensationpool.
H3b: The employer'sresidualprofit will be greaterwhen the employerhas discretion
over the allocationof the employee compensationpool than when the employer
does not have discretionover the allocationof the employee compensationpool.
Notwithstandingthese hypotheses, there are at least two behavioralforces working

againstthe predictions.First,as notedearlier,workplacenormsrelatedto cooperation,trust,
and reciprocitymay naturallyalleviate employerand employee opportunism,leaving little
role for compensationcontractsto mitigatesuch opportunism.Second, with respectto
Hlb,
H2b, and H3b, the effectivenessof discretionaryallocationto mitigateemployee opportun-
ism depends on whetheran employeruses her privateinformationin a mannerthat moti-
vates employee effort (B-R). While ideally the employerwill align an employee's compen-
sation with his individual performance,research suggests that employers are biased in
performanceevaluation(Prendergast1999). For example, employers may be unwilling to
make significantdistinctionsbetweenemployees'contributionsand compressthe individual
signals, leading to a centralitybias. Alternatively,employersmay make more extremedis-
tinctions between employees than is evident in their performancesignals, resulting in a
halo effect for higher-performing employees.
III. METHOD
Participants and Design
Two hundredthirty-sevenundergraduatestudents participatedin our computer-based
laboratoryexperiment,which employed a 2 (pool size discretion) x 2 (pool allocation
discretion) x 8 (period) mixed design.9 We manipulatedpool size discretion and pool
allocationdiscretionbetween-subjectsand periods(work sessions) within-subjects.We ran-
domly assigned participants to one of the four compensation contract conditions: discre-
tionary pool size/nondiscretionary pool allocation, discretionary pool size/discretionary
pool allocation, nondiscretionary pool size/nondiscretionary pool allocation, and nondis-
cretionary pool size/discretionary pool allocation.
" Two hundredforty-threestudentsparticipatedin the experiment;we eliminatedtwo groups (six participants)

due to computer-related
malfunctionsthatresultedin missingdata for some membersof these groups.
572 Fisher,Maines, Peffer and Sprinkle
Procedure
One-third(two-thirds)of the participantsassumedthe role of employer(employee) in
the experimentaltask. For each of the four compensationcontracts,the administratorran-
domly matched one employer with two anonymousemployees. The first few computer
screensexplainedthe productiontask, which involveddecodingnumbersinto letters(Chow
1983). Participantsthen received a descriptionof one of the four compensationschemes
tailoredto their experimentalcondition,along with severalnumericalexamples. Each em-
ployee's compensationcontracttook the following generic form in each period:
TPi = [a x (Ti - W,)] + B, (1)
where:
TP, = total pay for employee i, i = 1, 2;

a = pay per unit of leisure time consumed;
T, = total time availablefor employee i;
Wi = time allocatedto work by employee i; and
B, = work compensationfor employee i.
The employer'sresidualprofitin each periodwas computedaccordingto the following

generic formula:
RP = [r x (O, + 02)1 - (B + B2) (2)
where:
RP = residualprofitreceived by the employer;

r= revenueper unit of output;
+
0, 02 = outputof the two employees (groupoutput);and
B, + B2 = total work compensationfor the two employees.
In each of the four contracts,T, = 180 seconds, a = 0.5 francs, and r = 3.54 francs
(one franc,our experimentalcurrency,equaledone cent). For the two nondiscretionary pool
size contracts(NDS-NDA and NDS-DA), the total employee compensationpiece rate was
2.36 francsper unit of group output.
Our choice of parametervalues was driven by the determinationof a piece rate that
provides incentives for employees to shirk in the NDS-NDA contract.For employees to
want to shirk,the leisurepay per second mustexceed the compensationan employee would
receive for devoting an additionalsecond to work. We fixed leisure pay at a = 0.5 francs
per second. We then determineda piece rate for each employee of 1.18 francs per unit
based on: (1) pilot-testingin which participantsdecoded 58 numberson average in three
minutes (approximately0.32 numbersper second) and (2) a marginalper capita return
(MPCR) of 0.75 based on a common value used in public goods games (Ledyard1995).
Thus, an averageparticipantin the NDS-NDA contractcould earn either 0.5 francs from
The AccountingReview,April 2005
allocating one second to leisure or 0.38 (1.18 x 0.32) francs from allocatingone second
to work, leading to an incentiveto shirk.'"
The contractdescriptionvariedfor each conditionas follows.
* In the DS-NDA compensationcontract,Bi was describedas the total compensation
pool determinedby the employerdividedby two (i.e., split equally between the two
employees).
* The DS-DA compensationcontractwas describedby Equations(1) and (2), with a
note that the employerhad discretionto determineBi for each employee.
* In the NDS-NDA compensationcontract,Bi was described as 1.18 x (O, + 02),
where 1.18 francsequaledthe nondiscretionaryamountof individualemployee com-
pensation per unit of group output. This amountrepresentedan equal division of
revenueamong the three firm members(3.54/3). Thus, the total employee compen-
sation pool for the two employees was computedas 2.36 per unit of group output.
* In the NDS-DA compensationcontract,Bi was described as (d,) x (2.36) x (O,
+ 02) where equaled the fraction of the total compensationpool the employer
di
decided to give to employee i (d, + d2 = 1) and 2.36 francsequaledthe nondiscre-
tionaryamountof total employee compensationper unit of output.
The computerscreens then describedthe task. For each of eight work sessions, each
employee allocated 180 seconds between work and leisure. After every work session,
each employee learnedthe numberof items he decoded correctly.The employer received
a privatesignal for each employee, termed reportedoutput. Reportedoutput equaled the
actual numberof items correctlydecoded by the employee adjustedby a randomlychosen
numberbetween -7 and +7 using a uniformdistribution,and reflects the typical situation
in which employersreceive noisy measuresof individualemployee's performance."The
employerandemployees learnedtotal reportedgroupoutput(i.e., the sum of reportedoutput
for the two employees); however,only the employer knew reportedoutput for each indi-
vidual employee.
All participantscompleted three three-minutetrainingsessions to ensure they under-
stood the task. Additionally,prior to startingthe experimentalsessions, participantsan-
swered several manipulationcheck questions to ensure that they understoodthe incentive
contracts,the informationreceived by employees and employers,and the decisions to be
made by employees and employers.If a participantanswereda question incorrectly,then
the computerreturnedto the screencontainingthe appropriateinformation.Participantshad
to answer all questionscorrectlybefore proceedingto the work sessions.
After each work period, the computerautomaticallydeterminedthe pool size and/or
pool allocationin the nondiscretionaryconditions.The employerdeterminedthe pool size
and/or pool allocation in the discretionaryconditions. Employees' compensationand the
employer's residualprofitwere based on reported,ratherthan actual, outputbecause par-
ticipantslearnedtotal reported,but not total actual, output.After completingall eight pe-
riods, participantscompletedan exit questionnaireand received their compensation.
Jo As an example,if bothemployeesfully consumeleisure,each wouldearn90 francs(180 seconds x 0.5 francs)

in every work session. If two averageemployeesallocateall time to work in a session, each would earn 137
francs(58 units x 2 persons x 1.18 francs/unit).However,if one employeeengages in free riding,he could
earn 90 francsfrom leisureplus 68 francsif the otheremployeeworks for the entireperiod(58 units x 1.18
francs per unit), for a total of 158 francs. Thus, free riding is the dominantstrategywhen total employee
compensationis split evenly betweenthe two employees,althoughoutputis maximizedwhen both employees
work.
" Eachemployee'sreportedoutputcan be viewed as a functionof the numberof items he decodedcorrectlyand
a randomstateof nature.Reportedoutputwas constrainedto be greaterthanor equal to zero.
574 Fisher,Maines, Peffer,and Sprinkle
Dependent Measures
We examine three dependentmeasuresto test our hypotheses:employees' group per-
formance (numberof items actually decoded correctly),employees' group compensation,
and employer's residualprofit. We use actual, ratherthan reported,performanceas a de-
pendentmeasurebecause it reflectsemployee outputwithoutthe effect of the randomerror.
For employees' group compensation,we combine the correspondingamountsfor the two
employees pairedwith one employer(describedabove in Equation(1)) to reflectthe group's
total compensation.The employer's residual profit was calculated as shown in Equation
(2).
IV. RESULTS
Presentation of Experimental Results
We present our results in three tables, one for each dependentmeasure:employees'
group performance(actualoutput),employees' groupcompensation,and employer'sresid-
ual profit.12Panel A of each table presentsmeansover the eight periodsfor each dependent
measureby pool size discretionand pool allocationdiscretioncondition.Panel B of each
table containsthe relatedANOVA,with pool size discretionand pool allocationdiscretion
as the between-subjectsfactorsand periodas the within-subjects(repeatedmeasures)factor.
Additionally,Figure 1 presentsgraphicalresultsfor all eight periodsfor the threedependent
measures.
Employees' Performance (Hypotheses la and ib)

Table I and Panel A of Figure 1 presentresults for employees' actual group perform-
ance (output).Hypothesis la predictedthat employee groupperformancewould be greater
in the nondiscretionarypool size conditionsthan in the discretionarypool size conditions.
Hypothesis lb predictedthat employee group performancewould be greaterunder dis-
cretionary than nondiscretionarypool allocation. The results in Table 1 support these
predictions.
Panel A of Table 1 indicatesthat the mean performanceover all eight periodsfor the
nondiscretionarypool size conditionswas 111.08, comparedto mean performanceof 95.02
for discretionarypool size conditions(F = 9.90, p < 0.01, as shown in PanelB), supporting
Hia. As predictedby H1b, the mean performanceof 110.43 for the discretionarypool
allocation conditions was higher than 95.28 for the nondiscretionarypool allocationcon-
ditions (F = 8.64, p < 0.01).
The ANOVA in Panel B of Table 1 shows a statisticallysignificantmain effect for
period (F = 10.67, p < 0.01), a statisticallysignificantperiod-by-poolsize discretionin-
teraction (F = 8.04, p < 0.01), and a statistically significantperiod-by-poolallocation
discretioninteraction(F = 2.61, p < 0.02). As indicatedin Panel A of Figure 1, perform-
ance in the first period was roughlyequal across the four contracts(between 103 and 112
units); however, by the final period, there was significant variation in performance (a low
of 52 for the DS-NDA contract and a high of 117 for the NDS-DA contract). Thus, in the
12 Table 1 reports actual output, while Table 2 reports employee group compensation that is based on reported
output (actual output adjusted by a randomly selected number between -7 and +7). For this reason, group
compensation for the nondiscretionary pool size conditions (NDS-NDA and NDS-DA) in Table 2 does not equal
output in Table I multiplied by 2.36 francs.
FIGURE 1
Effect of Compensation Contracts on Employee Group Performance, Employee Group
Compensation, and Employer Residual Profit by Perioda
Panel A: Average Employee Group Performance by Periodb
130-
120
110
100 -
-*DS-NDA
- $A.- -
- DS-DA
90 - -- -
"$ . . ---&- - NDS-NDA
---- NDS-DA
80-
70-
60-
50
1 2 3 4 5 6 7 8
Period
Panel B: Average Employee Group Compensation by Periodc

295
275
255 -
235 - --- _
- 2-1-
----- +--DS-NDA
215 - ""
-. 1- 5- DS-DA
c195-
S175"- -- - NDS-NDA
15 5- --- NDS-DA
155 - \
135-
115-
95
,
1 2 3 4 5 6 7 8
Period
(continued on next page)
576 and Sprinkle
Fisher,Maines, Peffer,
FIGURE 1 (continued)
Panel C: Average Employer Residual Profit by Periodd
160 - .*
150 - '"- /
-
140 -- DS-NDA
-"--u-
8 130- -*- ----- - DS-DA
"- ?
-- -- NDS-DA
110 -
100 -
90 I I
, ,
1 2 3 4 5 6 7 8
Period
a The four contractsare
discretionarysize/nondiscretionaryallocation(DS-NDA),discretionarysize/
discretionaryallocation(DS-DA), nondiscretionary size/nondiscretionaryallocation(NDS-NDA),and
nondiscretionary size/discretionaryallocation(NDS-DA).
b Employeegroupperformance equalsthe sum acrossthe two employeesin each employer-employeegroupof
the actualnumberof items correctlydecodedby the employee.
c
Employeegroupcompensationequalsthe sum of the compensationreceivedby the two employeesin each
employer-employeegroup.
d Employer'sresidualprofitequals total revenueless compensationpaid to the two employeesin each employer-
employeegroup.
absence of explicit incentives, the effect of employee and employer opportunismhad a

greaterimpact on performanceas the employmentrelationshipapproachedits end.13
Employees' Compensation (Hypotheses 2a and 2b)

Table 2 and Panel B of Figure 1 presentresults for employees' group work compen-
sation (the amountof firm revenuereceived by employees).14 Hypothesis2a predictedthat
employee compensationwould be greaterin the nondiscretionarythan discretionarypool
13 An ANOVAfor reportedgroupperformanceshowedidenticalinferencesto those reportedin Table 1 for actual

group performance.Additionally,an ANOVAfor employee work time indicatedstatisticallysignificantmain
effects for pool size discretion,pool allocationdiscretion,period,the period-by-poolsize discretioninteraction,
and the period-by-poolallocation discretioninteraction.Similar to results for performance,opportunismin
employees'time allocationsbecamemore apparentin the laterperiods.In the firstperiod,employeesallocated
aboutthe same percentageof time to work in the worst (DS-NDA) contractand best (NDS-DA) contract(90
percentand94 percent,respectively).By the finalperiod,however,employeesin the DS-NDAcontractallocated
only 41 percentof theirtime to work,comparedto 90 percentfor the NDS-DA contract.
14 We also performedan ANOVAwith the dependentvariabledefined as employees' work compensationplus
leisure pay. This ANOVAmore completelyreflectsemployees'total benefits,i.e., the benefitsthat employees
receive from both theirwork effortand theiropportunism(time allocatedto leisure).Inferencesfrom ANOVA
resultsare the same as those reportedin PanelB of Table2 for work compensation.
TABLE 1
Employee Group Performance (Hla and Hlb)
Panel A: Means (Standard Deviations) for Employee Group Performance over All Periodsa
Compensation CompensationPool Sizeb
Pool
Allocationc Nondiscretionary Discretionary Overall
Nondiscretionary 105.92 85.18 95.28
(21.15) (26.20) (25.79)
n = 19d n = 20 n = 39
Discretionary 115.99 104.86 110.43
(14.80) (25.86) (21.55)
n = 20 n = 20 n = 40
Overall 111.08 95.02 102.95
(18.64) (27.56) (24.79)
n = 39 n = 40 n = 79
Panel B: ANOVAon Employee Group Performance

Source of Variation SS df MS Fr pf
Between-Subjects
Pool Size Condition (SC) 40,102 1 40,102 9.90 < 0.01
Pool Allocation Condition (AC) 34,982 1 34,982 8.64 < 0.01
SC x AC 3,645 1 3,645 0.90 < 0.35
Group(SC x AC) 303,674 75 4,049
Within-Subjects
Periodg 28,147 7 4,021 10.67 < 0.01
Period x SC 21,211 7 3,030 8.04 < 0.01
Period x AC 6,888 7 984 2.61 < 0.02
Period x SC x AC 2,937 7 420 1.11 < 0.36
Period x Group(SC x AC) 197,744 525 377
a
Employeegroupsconsist of two employeespairedwith one employer.Employeeactualgroupperformance
equalsthe sum acrossthe two employeesof the actualnumberof items correctlydecodedby the employee.
b We manipulated the determination of the compensationpool size at two levels: nondiscretionarilyset at 2.36
francs(cents) per unit of outputor discretionarilyset by the employer.In the nondiscretionary case, employees
in total receivedtwo-thirdsof revenue(3.54 per unit of output)and the employerreceivedone-thirdof
revenue.
SWe manipulatedthe allocationof the compensationpool betweenthe two employeesat two levels: a
nondiscretionary equal allocation(50 percentto each employee)or a discretionaryallocationby the employer.
d n = the numberof groupsin each treatment condition.
eThe Group(SCx AC) mean squareis the appropriate errortermfor the between-subjectseffects. The Period
x Group(SCx AC) meansquareis the appropriateerrortermfor the within-subjectseffects.
f We reporta one-tailedp-valuefor the maineffect of pool size and pool allocation,given the directional
predictionsfor these effects. All otherp-valuesare two-tailed.
g We manipulated Periodsas a within-subjectsfactorswith eight levels, 1-8. Participantsin all contract
conditionscompletedeight work sessions.
size conditions,while H2b predictedthat employee compensationwould be greaterin the

discretionarythan the nondiscretionarypool allocation condition. The results in Table 2
supportthese predictions.
Specifically,Panel A of Table2 indicatesthat the mean employee groupcompensation
over all eight periods was 261.23 for the nondiscretionarypool size and 198.60 for the
discretionarypool size (F = 16.07, p < 0.01 as shown in Panel B). Panel A also shows
578 Fisher, Maines, Peffer, and Sprinkle
TABLE 2
Employee Group Compensation (H2a and H2b)
Panel A: Means (Standard Deviations) for Employee Group Compensationover All Periods"
Pool
(50.46) (81.95) (78.00)
n = 19d n = 20 n = 39
Discretionary 273.39 225.77 249.58
(35.94) (91.71) (72.86)
n = 20 n = 20 n = 40
Overall 261.23 198.60 229.52
(44.85) (90.15) (77.70)
n = 39 n = 40 n = 79
Panel B: ANOVAon Employee Group Compensation

Source of Variation SS df MS F pr
Between-Subjects
SC x AC 34,106 1 34,106 0.89 < 0.35
Group(SC x AC) 2,862,249 75 38,163
Within-Subjects
Periodg 182,438 7 26,063 11.20 < 0.01
Period x SC 173,781 7 24,826 10.67 < 0.01
Period x AC 17,064 7 2,438 1.05 < 0.40
Period x SC x AC 15,398 7 2,200 0.95 < 0.48
Period x Group(SC x AC) 1,221,853 525 2,327
"Employeegroupsconsist of two employeespairedwith one employer.Employeegroupcompensationequals

the sum of the compensationreceivedby the two employees.
b We manipulated the determination set at 2.36
of the compensationpool size at two levels: nondiscretionarily
francs(cents) per unit of outputor discretionarilyset by the employer.In the nondiscretionarycase, employees
revenue.
cWe manipulatedthe allocationof the compensationpool betweenthe two employeesat two levels: a
nondiscretionary equal allocation(50 percentto each employee)or a discretionaryallocationby the employer.
d
n = the numberof groupsin each treatmentcondition.
e The Group(SCx AC) mean squareis the appropriate errortermfor the between-subjectseffects. The Period
x Group(SCx AC) mean squareis the appropriate errortermfor the within-subjectseffects.
'We reporta one-tailedp-valuefor the maineffect of pool size and pool allocation,given the directional
predictionsfor these effects. All otherp-valuesare two-tailed.
g We manipulatedPeriodsas a within-subjectsfactorswith eight levels, 1-8. Participantsin all contract
that the mean groupcompensationof 249.58 for the discretionarypool allocationcondition

was greater than that of 208.94 for the nondiscretionarypool allocation condition (F
= 6.51, p < 0.01).
Panel B of Table2 indicatesa statisticallysignificantmaineffect for period(F = 11.20,
p < 0.01) and a statistically significant period-by-pool size discretion interaction (F
= 10.67, p < 0.01). These interactionsare reflected in trends in compensationover the
eight periodsshown in Panel B of Figure 1. The statisticallysignificantperiod-by-poolsize
discretioninteractionreflects both decreases in group performance(output)and increases

in employers'opportunisticbehaviorin the discretionaryconditionsover time. On average,
employers in the discretionarypool size conditions paid total employee compensationof
59 percentof revenue, comparedto the predetermined67 percentpaid in the nondiscre-
tionarypool size conditions.However,as suggested by the period-by-sizediscretioninter-
action, the percentagepaid declined over time. For example, in period 1, employersin the
two discretionarypool size conditionson averageallocated62 percentof total revenueto
employees, while in period 8, they allocatedonly 51 percentto employees.
Employer's Residual Profit (Hypotheses 3a and 3b)

Hypothesis 3a predictedthat the employer's residual profit would be greater in the
nondiscretionarysize conditions than in the discretionarysize conditions. Hypothesis 3b
predictedthatthe employer'sresidualprofitwould be greaterin the discretionaryallocation
conditionsthanin the nondiscretionaryallocationconditions.Table3 and PanelC of Figure
I presentresults for the employer'sresidualprofit;these resultsdo not supportH3a but do
supportH3b.
As shown in PanelA of Table3, employersin the nondiscretionarypool size conditions
averagedcompensationof 130.65, comparedto 139.65 in the discretionarypool size con-
ditions. Contraryto H3a, the main effect of pool size discretionis not statisticallysignifi-
cant, as indicatedin Panel B (F = 1.50, p < 0.88). Panel A also indicatesthat the mean
employer compensationwas 142.25 in the discretionarypool allocation conditions and
127.98 in the nondiscretionarypool allocation conditions. This difference is statistically
significant(F = 3.68, p < 0.03), consistentwith H3b.
As indicated in Panel C of Figure 1, there is considerablevariationin employers'
compensationacross the four contractsand eight periods.This variabilityis reflectedin the
statisticallysignificantperiod-by-poolallocationdiscretioninteractionin Panel B of Table
3 (F = 2.55, p < 0.02). The results in Panel C of Figure 1 suggest that there is little
difference across contractsin employer's residualprofit until the last two periods, where
profitappearsgreaterin the conditionsin which the employerhad allocationdiscretion.We
examine this issue furtherin the supplementalanalysis below.
Supplemental Analyses
To providefurtherinsightson our results,we performedtwo additionalanalyses.First,
we reranall three ANOVAs using only periods 3-6 to eliminate both learningeffects in
periods 1-2 and end-game effects in periods 7-8. The ANOVAs for group performance
and employee compensationshowed statisticallysignificantmain effects for pool size dis-
cretion and pool allocationdiscretion,similar to results in Tables 1 and 2; however,there
were no mainor interactiveeffects relatedto periods.These analysesindicatethatthe effects
of pool size and allocationdiscretiondocumentedin Tables 1 and 2 are not simply learning
or end-gameeffects.
The ANOVA for periods 3-6 for employer's residual profit showed no statistically
significant main or interactive effects. An ANOVA for periods 1-6 had similar results.
Thus, it appears that the statistically significant effect of pool allocation discretion presented
in Table 3 arises from results in periods 7-8 and likely represents end-game effects. That
is, the employer's ability to differentially allocate compensation among employees appears
to benefit the employer primarily near the end of the employment relation.
Second, we performed an ANCOVA for our three dependent variables, adding the group
piece rate from the prior period as a covariate. This piece rate was set at 2.36 francs in the
two nondiscretionary size conditions, but varied based on the employer's choice in the
580 Fisher Maines, Peffer and Sprinkle
TABLE 3
Employer Residual Profit (H3a and H3b)
Panel A: Means (Standard Deviations) for Employer Residual Profit over All Periods"
Pool
(25.27) (41.87) (34.53)
n = 19d n = 20 n = 39
Discretionary 136.70 147.81 142.25
(17.97) (41.09) (31.80)
n = 20 n = 20 n = 40
Overall 130.65 139.65 135.20
(22.44) (41.77) (33.74)
n = 39 n = 40 n = 79
Panel B: ANOVAon Employer Residual Profit

Source of Variation SS df MS Fe pf
Between-Subjects
SC x AC 603 1 603 0.07 < 0.80
Group(SC x AC) 664,058 75 8,854
Within-Subjects
Periodg 17,201 7 2,457 1.42 < 0.20
Period x SC 16,467 7 2,352 1.36 < 0.23
Period x AC 30,881 7 4,412 2.55 < 0.02
Period x SC x AC 16,385 7 2,341 1.35 < 0.23
Period x Group(SC x AC) 908,688 525 1,731
a
Eachemployerwas pairedwith two employees.An employer'sresidualprofitequals total revenueless
employeecompensation.
b We manipulatedthe determination of the compensationpool size at two levels: nondiscretionarily set at 2.36
francs(cents) per unit of outputor discretionarilyset by the employer.In the nondiscretionary case, employees
revenue.
c We
manipulatedthe allocationof the compensationpool betweenthe two employeesat two levels: a
nondiscretionary equalallocation(50 percentto each employee)or a discretionaryallocationby the employer.
d
n = the numberof groupsin each treatmentcondition.
eThe Group(SCx AC) mean squareis the appropriate errortermfor the between-subjectseffects. The Period
x Group(SCx AC) mean squareis the appropriateerrortermfor the within-subjectseffects.
'We reporta one-tailedp-valuefor the maineffect of pool allocationgiven the directionalpredictionsfor this
effect. The p-valuefor the main effect of pool size equals (1 - a/2), given thatthe resultsare in the opposite
directionof thatpredictedin H3a. All otherp-valuesare two-tailed.
g We manipulated Periodsas a within-subjectsfactorswith eight levels, 1-8. Participantsin all contract
discretionaryconditions, with an average of 2.09 francs. If employees simply respondto

the size of the compensationpool in choosing theirworkeffortsand resultingperformance,
then it is possible that differencesin piece rates between the discretionaryand nondiscre-
tionarypool size conditionscould lead to our results,ratherthanemployerdiscretionversus
nondiscretion,per se. The ANCOVAtests for this possibility by relating employee per-
formanceto the effects of pool size and pool allocationdiscretionaftercontrollingfor any
effects of differencesin group piece rates.
The prior-periodpiece ratecovariatewas statisticallysignificantin all threeANCOVAs.
However,the inferences for effects of pool size and allocation discretiondid not change
between the ANOVAs in Tables 1-3 and the related ANCOVAs.In the ANCOVAsfor
performanceand employee compensation,the maineffects for pool size discretionand pool
allocation discretionwere statisticallysignificantat p < 0.01, as in Tables 1 and 2. The
ANCOVAfor employerresidualprofit showed a nonstatisticallysignificanteffect for size
discretionand a statisticallysignificantallocationdiscretion effect (p < 0.01), similar to
results in Table 3. Overall,these findings suggest that our results in Tables 1-3 reflectthe
effects of differencesin employerpool size discretion,as opposed to differencesin group
piece rates between the discretionaryand nondiscretionarypool size conditions.
V. SUMMARY AND DISCUSSION

In this paper,we investigateimplicationsof employerdiscretionover the total amount
and allocation of employee compensationfor firm output and employer/employeeremu-
neration.We find that a nondiscretionaryapproachto determiningtotal employee compen-
sation and a discretionaryapproachto allocatingemployee compensationincrease output
and employee pay. The employer's residualprofit increases with discretionover pool al-
location, but does not increase with a nondiscretionaryapproachto determiningtotal em-
ployee compensation.Additionally,our results indicate that employer and employee op-
portunismdo not completely destroy productiveoutcomes in the absence of mitigating
controls;however,the negative effects of opportunismincreaseover time, as the employ-
ment relationshipnears its end.
Our paperhas severalimplicationsfor the role of compensation-basedcontrolsystems
in mitigatingpotentialemployer and employee opportunism.First, our paper provides a
benchmarkfor the importanceof control systems by documentingthe effects of opportun-
ism in the absence of mitigatingcontrols.Our results show that employersand employees
do not engage in extremeopportunismin the absenceof controls,suggestingthatworkplace
norms lead employees to work and employersto rewardthis work to their mutualbenefit.
These findingshighlightthe need to comparethe costs and benefits of compensationsys-
tems not only with other control systems, but also with naturallyoccurringsocial mecha-
nisms associatedwith employmentsettings."5
Second, our results supportthe use of compensation-basedcontrol systems since some
opportunismexists in their absence. A compensationsystem based on nondiscretionary
compensationpool size and discretionarypool allocationincreasesfirmoutput.Employees
also receive greaterpay from such a compensationsystem;however,thatemployersdo not
profit from relinquishingdiscretion over the size of the compensationpool. This result
suggests that employers may decide to retain discretionover aspects of compensationto
the possible detrimentof total firm outputand employee compensation.However,compe-
tition among employers for employees could lead to employers relinquishingthe right to
set the compensationpool size.
' Some researchhas foundsynergiesbetweenformalcontrolsystemsand workplacenorms.Specifically,formal

controlsystemscan enhancethe trustthatservesas the foundationfor cooperativebehaviorin businesssettings
(Colettiet al. 2005).
582 Maines, Peffer and Sprinkle
Fisher,
Our paper also complementsresearchon the use of contractibleand noncontractible

informationin compensationcontracts.Ourempiricalresultssupportthe analyticalconclu-
sions of Baiman and Rajan (1995) regardingthe benefits of a contract in which total
employee compensationis based on contractibleinformationand the allocation of this
compensation is based in part on noncontractibleinformation.Since much of a firm's
contractibleinformationemanatesfrom the accountingsystem, our paperalso supportsthe
idea thatpubliclyavailableaccountinginformationand employer'sprivateinformationhave
complementaryroles in performanceevaluationand compensation.
Future researchcould provide additionalinsights on issues examined in this paper.
First, as indicatedearlier, we made specific choices for compensationparametervalues;
futureresearchcould examine the sensitivityof our resultsto our compensationparameter
values. Specifically,researchcould examine differentpiece rates and sharingrules among
employers and employees to determine if our results are robust across different
parameterizations.
Second, future researchcould examine other aspects of our design choices, such as
anonymoustwo-employee groups and use of one noisy signal. Employee shirkingcould
increase with largergroups of employees, strengtheningthe need for compensation-based
control systems.'6In contrast,the effectivenessof discretionover pool allocationmight be
reducedwhen employees are not anonymousif employerscompensateemployees on per-
sonal characteristicsunrelatedto performance(Ittner et al. 2003). Research also could
investigatethe efficacy of discretionaryallocationwhen employersmustarriveat subjective
evaluationsof employee performanceby aggregatingseveralperformancesignals. Research
has documentedbiases in combiningsuch information(Lipe and Salterio2000; Prendergast
and Topel 1993). Futureresearchcould examinewhetherthese biases overcomethe benefits
of the NDS-DA contractcomparedto the NDS-NDA contract.
Finally,futureresearchcould distinguishbetweeneffects of nondiscretionversuscom-
mitmentin determiningthe size of the compensationpool. In our experiments,employers
who did not have discretion were by definition committed to a compensationcontract.
However, employers can commit ex ante to a contractbased on contractibleinformation
and still retain discretion over characteristicsof the compensationcontract, such as its
functional form, specific parameters,and length. While we find that employers do not
benefit by relinquishingdiscretionover the size of the compensationpool, it is possible
that employersdo benefitfrom committingto a compensationarrangementover which they
have discretionto set the terms as suggested by Baiman and Rajan(1995). However,an-
alytic researchassumes that employers know employees' utility functions and, therefore,
can set contractterms optimally.In practice,the task of understandingex ante precisely
what motivatesemployees may be difficult.Thus, commitmentmay not benefitan employer
comparedto the situationof retainingfull discretion.Futureresearchon these issues will
increaseour understandingof how specific characteristicsof employmentcontractsmitigate
employee and employeropportunism.
REFERENCES
Axelrod, R. 1984. The Evolution of Cooperation. New York, NY: Basic Books.
Baiman,S., and M. V. Rajan. 1995. The informationaladvantagesof discretionarybonus schemes.
The Accounting Review 70 (October): 557-579.
'6 Some public good research, however, has found that individuals' contributions to the common pool actually
increase as group size increases (Issac and Walker 1988).
Berg, J., J. Dickhaut,and K. McCabe. 1995. Trust,reciprocity,and social history.Games and Eco-
nomic Behavior 10 (July): 122-142.
Bull, C. 1987. The existenceof self-enforcingimplicitcontracts.The QuarterlyJournalof Economics
102 (February):147-160.
Chow, C. W. 1983. The effects of job standardtightnessand compensationscheme on performance:
An exploration of linkages. The Accounting Review 58 (October): 667-685.
Coletti, A. L., K. L. Sedatole,and K. L. Towry.2005. The effect of controlon trustand cooperation
in collaborativeenvironments.TheAccountingReview80 (April):477-500.
Demski, J. 1997. Managerial Uses of Accounting Information. Boston, MA: Kluwer Academic
Publishers.
Fehr,E., S. Gichter, and G. Kirchsteiger.1997. Reciprocityas a contractenforcementdevice. Econ-
ometrica65 (July):833-860.
-, and - . 2000. Fairnessand retaliation:The economics of reciprocity.Journalof Eco-
nomic Perspectives 14 (Summer): 159-181
Fisher,J. G., S. A. Peffer,and G. B. Sprinkle.2003. Budget-basedcontracts,budgetlevels, andgroup
productivity. Journal of Management Accounting Research 15: 51-74.
Isaac,R. M., andJ. M. Walker.1988.The assuranceproblemin a laboratorymarket.QuarterlyJournal
of Economics 103 (July): 179-200.
Ittner,C. D., D. F. Larcker,and M. W. Meyer.2003. Subjectivityand the weightingof performance
measures:Evidencefrom a balancedscorecard.TheAccountingReview78 (July):725-758.
Ledyard,J. 0. 1995. Publicgoods: A surveyof experimentalresearch.In Handbookof Experimental
Economics,edited by H. H. Kagel, and A. E. Roth, 111-194. Princeton,NJ: PrincetonUni-
versityPress.
Lipe, M. G., and S. E. Salterio.2000. The balancedscorecard:Judgmentaleffects of common and
uniqueperformancemeasures.The AccountingReview75 (July):283-298.
Milgrom,P. R., and J. Roberts.1992. Economics,Organization& Management.UpperSaddleRiver,
NJ: PrenticeHall.
Prendergast,C., and R. Topel. 1993. Discretionand bias in performanceevaluation.EuropeanEco-
nomic Review37 (April):355-365.
.. 1999. The provisionof incentivesin firms.Journalof EconomicLiterature37 (March):7-
63.
Sprinkle,G. B. 2003. Perspectiveson experimentalresearchin managerialaccounting.Accounting,
Organizations and Society 28 (February-April): 287-318.
Zimmerman, J. L. 2003. Accounting For Decision Making and Control. Boston, MA: McGraw-Hill.

Performance 3

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Performance 3

Uploaded by

Copyright:

Available Formats

An Experimental Investigation of Employer Discretion in Employee Performance Evaluation and

We thankWaltBlacconiere,Dave Barrett,Daniel Beneish,Bob Bowen, Dave Greene,JohnHassell,Dan Heitger,

providesthe employerat least some discretionto allocatethis compensationusing

Throughoutthe paper,we use feminine pronounsto refer to employersand masculinepronounsto refer to

The AccountingReview,April 2005

Firms also have informalsources of information.This informationcan reflect multiple

The Accounting Review, April 2005

Finally, our study complements research on the use of contractible(public) versus

II. BACKGROUNDAND HYPOTHESES

The Accounting Review, April 2005

The Accounting Review, April 2005

contractwith respect to employer discretion/nondiscretionin determiningthe size of the

Compensation Contracts to Mitigate Employer and Employee Opportunism

Mitigating Employee Opportunism-Discretionary Pool Size/DiscretionaryPool

Mitigating EmployerOpportunism-NondiscretionaryPool Size/NondiscretionaryPool

The Accounting Review, April 2005

no formalcontrolsfor employee opportunism;however,the NDS-NDA contractdoes mit-

Mitigating Both Employeeand Employer Opportunism-NondiscretionaryPool Size/

Hlb: Groupoutputwill be greaterwhen the employerhas discretionover the allocation

7 Thereare severalotherdifferencesbetweenB-R and our paper.First,the B-R model is a single-periodmodel

H2a: Employees'compensationwill be greaterwhen the employerdoes not have dis-

Employer's Residual Profit

H3a: The employer'sresidualprofitwill be greaterwhen the employerdoes not have

Notwithstandingthese hypotheses, there are at least two behavioralforces working

" Two hundredforty-threestudentsparticipatedin the experiment;we eliminatedtwo groups (six participants)

TPi = [a x (Ti - W,)] + B, (1)

TP, = total pay for employee i, i = 1, 2;

The employer'sresidualprofitin each periodwas computedaccordingto the following

RP = [r x (O, + 02)1 - (B + B2) (2)

RP = residualprofitreceived by the employer;

The AccountingReview,April 2005

Jo As an example,if bothemployeesfully consumeleisure,each wouldearn90 francs(180 seconds x 0.5 francs)

The Accounting Review, April 2005

Employees' Performance (Hypotheses la and ib)

Panel B: Average Employee Group Compensation by Periodc

The Accounting Review, April 2005

Panel C: Average Employer Residual Profit by Periodd

absence of explicit incentives, the effect of employee and employer opportunismhad a

Employees' Compensation (Hypotheses 2a and 2b)

13 An ANOVAfor reportedgroupperformanceshowedidenticalinferencesto those reportedin Table 1 for actual

Panel B: ANOVAon Employee Group Performance

size conditions,while H2b predictedthat employee compensationwould be greaterin the

The Accounting Review, April 2005

Panel B: ANOVAon Employee Group Compensation

"Employeegroupsconsist of two employeespairedwith one employer.Employeegroupcompensationequals

that the mean groupcompensationof 249.58 for the discretionarypool allocationcondition

discretioninteractionreflects both decreases in group performance(output)and increases

Employer's Residual Profit (Hypotheses 3a and 3b)

Panel B: ANOVAon Employer Residual Profit

discretionaryconditions, with an average of 2.09 francs. If employees simply respondto

The Accounting Review, April 2005

V. SUMMARY AND DISCUSSION

' Some researchhas foundsynergiesbetweenformalcontrolsystemsand workplacenorms.Specifically,formal

The Accounting Review, April 2005

Our paper also complementsresearchon the use of contractibleand noncontractible

The Accounting Review, April 2005

The Accounting Review, April 2005

You might also like