Anda di halaman 1dari 16

Attribution Theory and Discipline Arbitration

Author(s): Brian Bemmels


Source: Industrial and Labor Relations Review, Vol. 44, No. 3 (Apr., 1991), pp. 548-562
Published by: Cornell University, School of Industrial & Labor Relations
Stable URL: http://www.jstor.org/stable/2524161 .
Accessed: 02/03/2014 11:15

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.

Cornell University, School of Industrial & Labor Relations is collaborating with JSTOR to digitize, preserve
and extend access to Industrial and Labor Relations Review.

http://www.jstor.org

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
ATTRIBUTION THEORY AND DISCIPLINE ARBITRATION

BRIAN BEMMELS*

Thisstudyfocuseson 230 malearbitrators' decisionsin a hypothetical


dischargegrievancecase. An analysisof the responsessupportsthe
central propositionof attributiontheory that a decision-maker's
responseto an individual'saction largelydepends on the decision-
maker'sattributions of causalityor responsibilityfor the action.The
analysisalso showsthat,all elseequal,thearbitratorstendedto specify a
more lenientpenaltywhen the grievantwas femalethan when the
grievantwas male; and moreexperiencedarbitrators tendedto decide
eithercompletely in favorof thegrievant or completely in favorof the
employer,whereas less experienced arbitratorstended to make
compromisedecisions.

M OST researchon grievancearbitra- arbitrators'decision-makingin discipline


tion has focused on legal processes arbitration.Attributiontheory is used to
and doctrines, and very little has been investigatethe sources of grievantgender
grounded in the behavioral sciences. Re- effects and the impact of arbitrators'
cently,several studies have found that the experience on theirdecisions.
outcomes of the arbitrationof discipline
grievances are affected by the grievants'
gender, and other studies have suggested AttributionTheory
thatthe personal characteristicsof arbitra-
tors are related to their decisions. But Attributiontheoryhas been developed
preciselyhow grievantgender and arbitra- to explain the behavior of decision-makers
tor characteristicsaffect arbitrationout- who must decide on a response to the
comes remains unknown. Legal doctrines behavior of other individuals, such as
are of little use in explaining these managers responding to the performance
phenomena. of their subordinates (for example, see
In this investigation,I utilize a behav- Green and Mitchell'1979) and officialsin
ioral theory in an experimental study of the criminaljustice systemresponding to
individuals alleged to have committedan
offense (Greenberg and Ruback 1982).
* The author is Associate Professorof Industrial Attributional processes are a form of
Relations at the Faculty of Business, Universityof cognitive information processing that
Alberta.He thanksRay Rasmussen for many helpful helps decision-makersto categorize infor-
suggestionson this research project, E. G. Fisher,
Yonatan Reshef, Kay Stratton-Devine,and C. Brian mation, reduce ambiguity, and under-
Williams for comments on an earlier draft of this stand the behavior of others. Mitchell
paper, and Caroline Taylor for research assistance. (1982) notes that attributiontheoryis one
Copies of the case transcript,questionnaire,data, of the most prominentand activeresearch
and computerprogramsused to generate the results
of this study are available from Brian Bemmels,
topics in social psychology and related
Facultyof Business, Universityof Alberta, Edmon- subjects.
ton, Alberta,CANADA T6G 2R6. Attribution theory is well suited for

Industrialand LaborRelationsReview,Vol. 44, No. 3 (April 1991). ? by Cornell University.


0019-7939/91/4403$01.00

548

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
ATTRIBUTION THEORY AND DISCIPLINE ARBITRATION 549

studying the arbitration of discipline determine the decision-maker's attribu-


grievances,since in theirdecisions arbitra- tions of causality for the individual's
tors are respondingto the behavior of the behavior; second, the decision-maker's
grievant in the case. A first step in causal attributionsdetermine the appro-
disciplinearbitrationis to decide if thereis priate response to the individual's behav-
adequate evidence that the alleged inci- ior. The causal attributionsare an inter-
dent or behavior occurred. Attribution mediate step in the decision-making
theory is not applicable to this decision, process, and facilitate a connection be-
whichis based on the factsconcerningthe tween the facts of the case and the
incident apart from any consideration of decision. By substitution,in the reduced-
who may have caused the incident. In form relation, the facts of the case
many discipline arbitration cases, how- determinethe final decision.
ever, the occurrence of the incident or Kelly (1967) argues that the formation
behavior that led to the discipline is not of causal attributions(that is, the first
disputed, and the arbitrator'sprimary(or structuralrelation) is governed primarily
sole) duty is to decide if the penalty by information relating to three vari-
assessed by the employeris appropriate- ables-consensus, distinctiveness, and con-
a decision that mightwell be explained by sistency.Consensus is the extent to which
attributiontheory. otherindividualsin the same environment
The basic premise of attributiontheory exhibit the same behavior. If many other
is thatthe decision-maker'sresponse to an individuals exhibit the same behavior
individual's behavior will depend in large (high consensus), causality is likely to be
part on how the decision-makerviews the attributedto the environment.If no other
causes of the behavior. Attributionre- individualsexhibitthe same behavior (low
search has shown that causal attributions consensus), causalityis likelyto be attrib-
vary on an internal-externaldimension uted to the individual. Thus, consensus
relating to the locus of control. An relates to variation in behavior across
internal causal attributionis made when individuals,all else held constant.
the decision-makerbelieves factorswithin Distinctivenessis the extentto whichthe
the individual'scontrol caused the behav- behavior in question is unique to a specific
ior in question, and thereforethe individ- entity.For example, suppose an employee
ual is responsible for the behavior. An has missed a deadline for completing a
external causal attributionis made when project (entity). If this employee has
the decision-makerbelieves somethingin frequently missed deadlines on other
the environmentor circumstancesbeyond projects (low distinctiveness),causality is
the individual'scontrol caused the behav- more likelyto be attributedto the individ-
ior. ual (the employee does not plan ahead or
The decision-maker'scausal attributions does not considerdeadlines important).If,
will influence his or her response. If the however, this employee has always met
cause of the behavior is attributedto the deadlines on other projects (high distinc-
individual (such as lack of effortcausing tiveness),then causalityis more likelyto be
poor job performance),then a response attributedto the situation (this deadline
directlyaffectingthe individual (such as a was unreasonable). Thus, distinctiveness
reprimand) is appropriate. If, however, relates to variation in behavior across
somethingin the individual'senvironment entitiesfor the same individual.
is determined to be responsible for the Consistencyis the extent to which the
behavior (such as poor management or behavior in question has been exhibitedby
equipment failures), then a response af- the individual on this entityin the past.
fectingthe individual is not appropriate For example, suppose the deadline the
(Lord and Smith 1983). employee has missed is for a routine
Thus, the attributionmodel is a recur- project. If this employee has frequently
sive (or stepwise) systemwith two struc- missed this deadline in the past (high
tural relations. First,the factsof the case consistency),then causalityis more likely

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
550 INDUSTRIAL AND LABOR RELATIONS REVIEW

to be attributedto the individual. If this The followingdiscussion of moderators


employee has never missed this deadline will be limitedto the personal characteris-
in the past (low consistency),then causality ticsof the decision-makerand the individ-
is more likely to be attributed to the ual, specifically,the experience of the
environment(othercircumstancesprohib- arbitratorand the gender of the grievant.
ited meeting the deadline this time).' These moderators are chosen for discus-
Thus, consistencyrelates to variation in sion because research has found that they
behaviorover timeforthe same individual influence arbitrators' decisions (Nelson
and the same entity. and Curry 1981; Bemmels 1991).
Information pertaining to these three In the attributionmodel, moderators
variables-consensus, distinctiveness,and could enterthe decision-makingprocess in
consistency-isoftenmade available to the several ways. They could influence the
arbitrator.Informationrelatingto distinc- firststructuralrelation,that is, theycould
tiveness and consistencywill be revealed influence the formationof causal attribu-
by the grievant'sprior disciplinaryrecord. tions. For example, studies of supervisors'
For example, prior disciplinaryactionsfor attributionsof causality for job perfor-
related incidentswill indicate high consis- mance have found that supervisors are
tency,and prior discipline for unrelated more likelyto attributesuccess to external
incidentswill indicate low distinctiveness. causes and failureto internalcauses when
If the behavior that led to the discipline they are evaluating women than when
under dispute is common among other they are evaluating men (Deaux and
employees,indicatinghigh consensus,this Emswiller 1974; Dobbins 1985; Feldman-
fact may be pointed out by the union in Summers and Kiesler 1974; Garland and
defense of the grievant. Price 1977; Nieva and Gutek 1980; and
Rose 1978). If a gender bias occurs in the
Moderators in AttributionTheory
causal attributionsof arbitrators,this bias
Other factors,termed moderatorshere, could explain the gender effectsthatwere
may affect the attributional decision- found in the arbitration studies noted
making process. Green and Mitchell above (see Heilman and Guzzo 1978).
(1979) discuss a number of potential Kelly (1972) has also argued that deci-
moderators, such as self-servingattribu- sion-makersdevelop "causal schemata" or
tions,the seriousnessof the behavior, the simplifiedrules to help them formcausal
relationship(if any) between the decision- attributions. Causal schemata may be
maker and the individual, and the per- developed from experience by observing
sonal characteristicsof the decision-maker cause and effect relations or may be
or the individual (also see Kelly and borrowed frompopular wisdom about the
Michela 1980). Moderatorsmaybe sources causal structure of certain behaviors.
of error or bias in the decision-making These schemata may be importantwhen
process, or they may be legitimatecon- relevantinformationis not available. They
cerns that the decision-maker should allow the decision-maker to reach an
consider.2 acceptable causal attributionby relyingon
a common causal scheme and the informa-
1
Althoughpoor job performanceis used here for tion that is available, and inferringthe
illustration,attributiontheorycan easily be applied missinginformationto fill out the causal
to otherbehaviorsthatcommonlylead to disciplinary scheme. Gender effects in arbitrators'
action. causal attributions might indicate that
2 In addition to moderatorsin attributiontheory,
Green and Mitchell (1979) also discuss boundary
constraintsaffectingthe attributiondecision-making
process. Boundary constraintsare limitationson the pronounce another decision by the collectiveagree-
applicabilityof causal attributionsto the decision. ment. In the experimentalstudydiscussed below, the
The collective agreement may be an important clauses fromthe collectiveagreementincluded in the
boundary constraint in grievance arbitration. For case were writtento avoid boundary constraintson
example, an arbitratormay arrive at one decision the applicabilityof the arbitrators'causal attributions
based on causal attributions,but be constrained to to theirdecisions.

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
ATTRIBUTION THEORY AND DISCIPLINE ARBITRATION 551

arbitratorshave differentcausal schemata the cognitive processes underlying each


for men and women, or that gender is a structuralrelation.
key piece of informationused to infer
other information in their causal sche- Methods
mata. If arbitrators'experience moderates
their causal attributions,it may be that The present studyis a field experiment
experience has changed theircausal sche- that teststhe attributiontheoryof arbitra-
mata. tor decision-makingin discipline cases. A
Moderators could also influence the transcriptof a discharge case was mailed
second structuralrelation in the attribu- to a sample of arbitrators.The arbitrators
tion model, that is, they could affectthe were asked to read the case and complete
decision given the causal attributionsof a questionnaireindicatinghow theywould
the decision-maker.For disciplinearbitra- decide the case and other information
tion,grievantgender effectsin the second necessary to test the theory. Facts of the
structuralrelationwould indicatea double case were manipulated to test attribution
standard in terms of the appropriate theory.The grievant'sgender and arbitra-
penalty for a given behavior. That is, tor'sexperience were testedas moderators
given the arbitrators'attributionsof cau- in the attributionaldecision-makingpro-
salityfora certainbehavior,the appropri- cess.
ate penaltywould depend on the gender
The Case
of the grievant.If arbitrators'experience
moderates the second structuralrelation, The case used in thisstudywas adapted
it follows that arbitrators' standards in from an actual discharge arbitrationcase.
terms of the appropriate penalty vary In experimentalstudies such as this one,
according to theirexperience. in which arbitratorsmust make a decision
Moderatorscould also enter each of the based on a transcriptof the case without
structuralrelationsin two ways, either as the advantage of a hearing, it is very
main effectsor as interactioneffects.Main importantto ensure that the case clearly
effectsalone would indicate a direct bias presents sufficientinformationto enable
on the causal attributionsor the final the arbitratorsto make a sound decision.
decision,but no effecton the processes of The transcriptincluded background in-
formingcausal attributionsfromthe facts formationabout the case, edited testimony
of the case or reachinga decision fromthe fromthe hearing, and relevantprovisions
causal attributions. Interaction effects, from the collective agreement. The case
however,would indicatethatthe processes was previewed by several experienced
are altered. Interaction effects between arbitrators (including a member of the
the moderatorsand the factsof the case in National Academy of Arbitrators) and
the firststructuralrelationwould indicate several industrial relations professors to
that the process by which the factsof the ensure thatthe informationwas appropri-
case are used by the arbitratorsin forming ate, clearly presented, and sufficientto
causal attributions (that is, the weight allow an informeddecision.
assigned to the various facts)is altered by The case involved an employee in the
the grievant's gender or the arbitrator's composing room of a daily newspaper.
experience. Interaction effects with the The grievant'sjob was to "lay out" the
causal attributionsin the second structural Classified Advertisement section. The
relationwould indicatethatthe process by grievant was discharged after failing to
which the arbitrators'attributionsof cau- place an 8-by-12-inch display advertise-
salitydetermine the appropriate decision ment in the newspaper. The composing
is altered. Thus, main effects would room supervisortestifiedthat he checked
indicate that the moderators produce a the computer files and found that the ad
shift in the outcome of each structural had been included in the ProductionPrint
relation,whereas interactioneffectswould Out. When asked about the ad during the
indicate a more fundamental change in supervisor's investigation prior to the

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
552 INDUSTRIAL AND LABOR RELATIONS REVIEW

discharge,the grievant'sresponse was that ment) from the paper. In the high-
"it must have gotten lost somewhere." consensus versionof the case, the supervi-
There was no dispute that the grievant sor statedunder cross-examinationthathe
had omittedthe ad. was aware of several other incidents in
The testimonyand evidence in the case which other employees had omitted an
focused on a dispute about whether item fromthe paper.
management (external causes) or the Distinctiveness.
Distinctivenessis the ex-
grievant(internalcauses) was responsible tent to which the behavior that led to the
for the incidentthat led to the discharge. discharge is the only behavioral problem
The Company argued that the omission with the grievant. In the low-distinctive-
was due to negligenceand carelessnesson ness versions of the case, the composing
the part of the grievantand there was just room supervisor testifiedthat on both of
cause for discharge. The union argued the grievant'stwo performanceappraisals
that laying out the Classified Advertise- since joining the company he had rated
ment section each day was an unreason- the grievant's overall performance as
able amount of work for one person, that "needs improvement,"and had pointed
management had been notified of this out specific respects in which improve-
staffingdeficiencyon severaloccasions but ment was needed. In the high-distinctive-
failed to take appropriate action, and the ness version, the supervisor testifiedun-
grievantshould not be punished for poor der cross-examinationthat he had rated
management. Except for the variations the grievant'sperformanceas "good," but
described below, all of the arbitrators had pointed out specificways in which he
received identicaltranscripts. or she could improve.
Consistencyis the extent to
Consistency.
which the grievanthas exhibitedthe same
Manipulations
behavior in the past. In the low-consis-
Three manipulationswere made in the tencyversions of the case, the composing
factsof the case to representhigh and low room supervisor testified under cross-
consensus,distinctiveness,and consistency. examination that the grievant had never
The factsaffectedby these manipulations before omittedan itemfromthe paper. In
werediscussedin themainbodyof thetran- the high-consistencyversion, the supervi-
scriptand were repeated in the closing ar- sor testifiedthatthe grievanthad received
guments of either management or the a writtenwarning (submittedas evidence)
union. In addition,the gender of the griev- approximately three months earlier for
ant was manipulated to test for grievant's omittingan In Memoriam ad from the
gender as a moderatorin the attributional paper.
decision-making process.Consequently,the Gender.Half of the cases identifiedthe
experiment is a 2 x 2 x 2 x 2, fully grievantas male (John) and half as female
crossedfactorialdesign(Campbelland Stan- (Karen). The grievant'sname appeared 6
ley 1963), with 16 cells and 16 different times, and pronouns (such as 'he" and
versionsof the case. The cases were ran- "him" or "she" and "her") referringto the
domlydistributedamong thearbitratorson grievant appeared 21 times throughout
themailinglist,withapproximatelyan equal the transcript.
numberof each versionof the case mailed
to the arbitrators. The Sample
Consensus.Consensus is the extent to
which other employees exhibit the same The arbitrationcase was mailed to 580
behavior as the grievant. In the low- arbitratorslisted in the National Academy
consensus versions of the case, the com- of Arbitrators'1986/87 Membership Direc-
posing room supervisortestifiedthatnone toryor the Industrial Relations Research
of the other 11 people employed in the Association's 1987 MembershipDirectory.
composing room had ever omitted a The latter included only IRRA members
necessaryitem (such as a paid advertise- who indicated arbitrationas their princi-

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
ATTRIBUTION THEORY AND DISCIPLINE ARBITRATION 553

pal occupation. To encourage participa- Table 1. Descriptive Statistics.


tion in the study,a summaryof the results Variablea Frequency Percent
was returned to arbitratorswho partici-
Causal Attributions
pated in the study. Two weeks after the
MRESP jgsb 93c
cases and questionnaires were mailed, a
GRESP 4.09b .83c
follow-upreminderwas mailed.
Thirty-twoof the cases and question- Decisions
naires were returned as undeliverable, SUSTAIN
Yes 168 73.0
leaving 548 successfulmailings.Responses No 62 27.0
were received from 290 arbitrators,for a AWARD
response rate fromthe successfulmailings Full Reinstatement 35 20.8
of 52.9%. Thirty-eightarbitrators indi- Partial Reinstatement 133 79.2
cated they were unwilling or unable to SUSPENSION (in days) 55.42b 49.97c
participate in the study; five forms were FactsoftheCase
unusable due to ambiguous or incomplete CONSEN
responses; and responses from the seven- High 108 47.0
teen arbitratorswho were female were Low 122 53.0
excluded from the study because they DIST
were too fewto controlforthe arbitrator's High 122 53.0
gender in the empirical analysis.3Thus, Low 108 47.0
the responses that were used were from CONSIS
High 127 55.2
230 arbitrators,all male. Low 103 44.8
Moderators
Variables GENDER
Male 116 50.4
Table 1 gives descriptive statisticsfor Female 114 49.6
the variables in the study. Two variables EXP (in years) 18.90b 11.15C
reflectthe arbitrators'causal attributions: a MRESP= extent to which the arbitrator deems
MRESP indicates the extent to which the managementresponsibleforthe incidentcausing the
arbitratorsbelieved management was re- discharge; GRESP= extent to which the arbitrator
deems the grievantresponsible;CONSEN = dummy
sponsible (externalcauses) forthe incident variable indicatingwhetherthe arbitratorreceived a
that led to the discharge, and GRESP high- or low-consensusversion of the case (see text
indicatesthe extentto whichtheybelieved for explanation); DIST=dummy variable indicating
the grievant was responsible (internal whether the arbitrator received a high- or low-
distinctivenessversion of the case; CONSIS= dummy
causes). These variables, which are from variable indicatingwhetherthe arbitratorreceived a
two correspondingitems on the question- high-,or low-consistencyversion of the case; GEN-
naire, are on a scale from 1 (not at all) to 5 DER = gender of the grievant; EXP = the arbitrator's
(entirely). The means indicate that the experience, in years.
arbitrators,on average, believed the griev- Mean.
c Standard deviation.
ant was primarilyresponsible, but that
management shared some responsibility
for the incident.4
totallyresponsibleat one end and the grievanttotally
responsible at the other end could indicate that the
3Controlling for the arbitrator'sgender is neces- arbitratorfelteitherthatboth were equally responsi-
sary when studying grievant gender effects in ble or that neither was responsible for the incident
arbitration,since grievantgender effectshave been (the arbitrator might attribute the incident to
found for male arbitratorsbut not female arbitrators coworker interference,bad luck, or chance, for
(Bemmels 1988c, 1990b). example). These two possible determinationsregard-
4MRESP and GRESP are included as two separate ing who was responsiblefor the incidentwould most
variables rather than as opposite ends of one likelylead to differentdecisions by the arbitratoron
continuum, since a single measure of responsibility the appropriatenessof penalizing the grievant.The
could not adequately capture and distinguish be- two scales used will adequately capture and distin-
tween all relevant possibilities. For example, a guish betweenthese two possibilities.The correlation
midpoint response on a scale with management between GRESP and MRESP is -.505.

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
554 INDUSTRIAL AND LABOR RELATIONS REVIEW

Three variables reflectthe arbitrators' in the presentcase, the manipulationsare


decisions. SUSTAIN indicates whether the not noted by the arbitrators,it is not likely
grievance was sustained or denied. The that theywill influencethe causal attribu-
grievance was sustained by 168 of the tions or decisions as theoreticallypre-
arbitratorsand denied by 62.5 The second dicted.
decision variable (AWARD) is the award Three items were included on the
named by the 168 arbitratorswho sus- questionnaire asking the arbitratorsabout
tained the grievance. Thirty-fivearbitra- the factsof the case relatingto consensus,
tors reinstated the grievant with full distinctiveness, and consistency. These
backpay, and 133 reinstatedthe grievant items (all on 5-point scales) measure the
but substituteda suspension as a lesser arbitrators'perceptionsof the factsof the
penalty.The thirddecision variable is the case and are used as manipulationchecks.7
lengthof SUSPENSION (in days) imposed by The mean difference on the question
the 133 arbitrators who reinstated the relating to consensus between arbitrators
grievantbut substituteda suspension as a who received the high-and low-consensus
lesser penalty. Fifty-sevenof the arbitra- versionsof the case is .96 (t = 10.90), and
tors imposed a suspension equal to the the comparable mean differences for
entiredelay fromthe date of the discharge distinctivenessand consistencyare .98 (t
untiltheirdecision. These suspensionsare = 9.36) and 1.09 (t = 10.56), respectively.
coded as 112 days, which is estimated as All threedifferencesare statisticallysignif-
the average delay from the date of the icant and in the expected direction.Thus,
dischargestatedin the case to the date the the manipulationsof the factsof the case
case was received by the arbitrators. apparentlywere successful.
Three dummy variables represent the A manipulationcheck was not included
factsof the case that were manipulated in for the grievant'sgender manipulation in
the experimentaldesign. CONSEN indicates order not to give the arbitratorsa clue that
whetherthe arbitratorreceived a high-or gender was an aspect of the study. Since
low-consensusversionof the case, DIST in- the grievant'sname and pronouns refer-
dicates whetherthere was high or low dis- ring to the grievantappeared frequently
tinctiveness,and CONSIS indicateswhether throughout the transcript,it is highly
there was high or low consistency.Two unlikely that the grievant's gender was
moderator variables are included. GENDER unnoticed by the arbitrators.
indicates the grievant's gender as identi-
fied in the transcript.The questionnaire
also asked each arbitratorhow manyyears EstimationProcedures
experiencehe had as an arbitrator,and EXP
is the second moderatorvariable.6 Since attribution theory is a simple
recursive model, assuming normally dis-
tributed error terms allows both the
Manipulation Checks structuraland reduced-formequations to
be estimatedby single equation methods.
In experimental studies it is important The structuralequations are the causal
that the manipulationsbe successful,that attributionsregressed on the facts of the
is, thatthe informationcues relatingto the case and the decision variables regressed
manipulationsbe noted by the subjects.If, on the causal attributions.The reduced-
form equations are the decision variables
5 The fact that most arbitrators believed the regressed on the facts of the case. The
grievantwas primarilyresponsible for the incident,
but the majority also sustained the grievance,
indicates that most of the arbitrators felt that 7For example, the item used as a manipulation
dischargewas too severe in thiscase. check for the consensus manipulation asked the
6 The experience variable reflectsonly one dimen- arbitrators,"Do you feel thatomittingitemsfromthe
sion of arbitrators'experience, how long they have paper is a common problem withother employees?"
been practicingarbitrators.It does not reflectthe The 5 response categoriesranged from"not at all" to
number of cases theyhave decided. "verycommon."

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
ATTRIBUTION THEORY AND DISCIPLINE ARBITRATION 555

moderatorvariablesare included in all the Table 2. FactorsAffecting


Arbitrators'
Causal
equations, since in attributiontheorymod- Attributions
in a Hypothetical
Discipline
erators can affecteither of the structural Arbitration
Case: RegressionAnalysis.
relations. Ordinary least squares is used (N = 230 Male Arbitrators;
t-Statistics
for the regressionson the causal attribu- in Parentheses)
tionsand SUSPENSION. For the two dummy Variablea MRESP GRESP
dependent variables,SUSTAIN and AWARD,
logit regressionis used. Constant 2.203*** 3.939***
(13.65) (25.02)
Three sets of interactionsamong the
independent variables were tested in the FactsoftheCase
estimation procedures, First, interactions CONSEN .527*** -.396***
among the variablesrepresentingthe facts (4.51) (- 3.74)
- .244** -.012
of the case (CONSEN X DIST, DIST X DIST
(- 2.09) I_ 11)
CONSIS, and CONSEN X CONSIS) or the
CONSIS -.321*** .267**
attributions(MRESP X GRESP) were tested.
(-2.76) (2.53)
These interactionscould provide informa-
tion about the arbitrators'causal schemata. Moderators
As it happens, however,these interactions GENDER -.039 -.056
were statistically insignificantin all of the (-.33) (-.85)
regressions,and theyare thereforeomit- EXP -.007 .010**
(-1.40) (2.06)
ted from the models reported below.
Second, interactionsbetween each moder- R-squared .13 .10
ator variable and the factsof the case or F forGENDER Interactionsb .94 .87
the attributionswere tested. These were F forEXP Interactionsb .74 .13
also statisticallyinsignificantin all of the For definitions,
a see note'a' to Table 1, or text.
b
regressions. Finally, an interaction be- F-ratioforinteractions withthefactsof thecase.
tween the two moderators(GENDER X EXP) ** Significant at the .05 level;***at the .01 level
(2-tailedtests).
is included in the regressionswhere it is
statistically
significant.
decision variables are presented in
AWARD
Table 3. Models 1 and 3 are the structural
Results
equations for the second link in attribu-
Table 2 presents the regression results tion theory for the SUSTAIN and AWARD
for the structuralequations representing variables,respectively.Models 2 and 4 are
the firstlinkin the attributionmodel. The the reduced-formequations.
coefficientestimateson CONSEN and CON- In Model 1, both MRESP and GRESP are
sis are statisticallysignificant with the statisticallysignificantwith the expected
expected sign for both MRESP and GRESP. signs,indicatingthatthe arbitrators'causal
The estimatein DIST, however,is insignif- attributionsdid influence their decisions
icant in the GRESP regression, and is to sustain or deny the grievance. GENDER,
significantbut witha sign opposite of what EXP, and all of the interactions are
is expected in the MRESP regression. statisticallyinsignificant.
GENDER is not statisticallysignificantin In Model 2, the reduced-formequation
either regression. EXP is positive and forthe SUSTAIN decision variable,all of the
statisticallysignificantin the GRESP regres- coefficientestimates on the facts of the
sion, indicating that more experienced case are statisticallysignificantwith the
arbitratorswere more likely to attribute expected sign. All of the interactionsare
responsibility for the incident to the insignificant.
grievant.The F-ratiosfor the GENDER and As in Model 1, GENDER is not significant
EXP interactionswith the factsof the case in Model 2. The estimateon EXP, however,
are insignificantand the GENDER X EXP is negative and statisticallysignificantin
interactionis also insignificant. Model 2, indicatingthatmore experienced
The logit results for the SUSTAIN and arbitratorswere less likely to sustain the

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
556 INDUSTRIAL AND LABOR RELATIONS REVIEW

of WhethertheGrievanceWas Sustainedand theAward.


Table 3. LogitAnalysis
(Z-Valuesin Parentheses)
SUSTAIN AWARD

Variablea 1 2 3 4
Constant 3.518** 1.838*** - 2.572* - 2.409***
(2.30) (5.39) (-1.75) (-4.50)
CausalAttributions
MRESP .606** .705***
(2.55) (2.67)
GRESP -.793*** -.328
(-2.84) (-1.17)
FactsoftheCase
CONSEN .545*** 1.046***
(3.24) (3.85)
DIST .345** -.143
(2.15) (-.62)
CONSIS -.418** - 744***
(-2.51) (-3.13)
Moderators
GENDER -.200 -.223 1.457*** 1.668***
(- 1.22) (- 1.39) (3.06) (3.24)
EXP -.021 -.033** .035* .026
(-1.42) (-2.34) (1.75) (1.25)
EXP*GENDER -.041** -.049**
(-2.12) (-2.32)
Log of Likelihood Function -112.66 -119.62 -69.97 -63.96
X2for GENDERinteractionSb 3.06 2.32 1.09 2.28
X2fOrEXP InteractionSb .20 1.38 1.26 3.78
Sample Size 230 230 168 168
a For definitions,see note 'a' to Table 1, or text.
b
Likelihood ratio X2for interactionswiththe causal attributionsor the factsof the case.
* Significantat the .10 level; ** at the .05 level; *** at the .01 level (two-tailedtest).

grievance.Comparing Model 2 to Model 1 equation (Model 2) reflectsthe combined


(that is, comparing the reduced-form experience effectfrom both of the struc-
equations to the structuralequation for tural equations.
the second link in the attributionaldeci- In Model 3, AWARD is regressed on the
sion-making process), the EXP estimate causal attributionsand the moderators.
declines by approximatelyone-third.Also, The estimate on MRESP is statistically
recall that EXP had a significanteffecton significantwiththe expected sign,but that
the arbitrators'attributionsof causalityto on GRESP is insignificant.The estimateon
the grievant in Table 2 (that is, in the grievant's gender is positive and statisti-
structuralequation for the firstlink of the cally significant,indicating that the arbi-
attributional decision-making process). tratorswere more likely to award a full
Together, these resultssuggest that about reinstatementrather than a partial rein-
one-thirdof the effectof the arbitrators' statementif the case identifiedthe griev-
experience on theirdecision to sustainthe ant as female rather than male. The
grievanceoccurs in the firststructurallink estimate on EXP was also positive and
of the attributionaldecision-makingpro- significant.The GENDER X EXP
statistically
cess, and about two-thirdsoccurs in the interaction is negative and statistically
second structural link. The estimated significant, indicating that the gender
experience effect in the reduced-form effect diminishes for more experienced

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
ATTRIBUTION THEORY AND DISCIPLINE ARBITRATION 557

arbitrators.The relativemagnitudesof the structurallink of the attributiondecision-


GENDER and GENDER X EXP parameter making process.
estimatesindicate that the gender effectis The regression results for the SUSPEN-
positive for inexperienced arbitrators,di- SION decision variable are presented in
minishesto zero at approximately35 years Table 4. Models 1 and 2 are estimates
of experience (1.457/.041), and becomes using the subsample of all the arbitrators
negativeforarbitratorswithmore than 35 who substituteda suspension in place of
years of experience. the discharge (n = 133). Models 3 and 4
In Model 4, the reduced-formequation
use a smaller subsample (n = 76) of only
for the AWARD decisionvariable,CONSEN the arbitratorswho substituteda suspen-
and CONSIS are statistically The
significant.
sion thatwas not equal to the lengthof the
pattern of results for the moderator
variables is similar to that in Model 3, delay fromthe date of the dischargeto the
except that EXP is insignificant. The date of theirdecision. This lattersubsam-
GENDER and EXP X GENDER estimates ple includes only the arbitratorswho, in a
change very littlebetween the structural sense, exercised greater discretion in
equation (Model 3) and the reduced-form choosing the appropriate length of sus-
equation(Model 4), and GENDER does not pension. In this subsample the average
have a significanteffecton the arbitrators' suspension was 12.9 days (s.d. = 11.7).
causal attributionsin Table 2, indicating Models 1 and 3 are estimates of the
that all of the gender effecton the AWARD structuralequation for the second link in
decision variable occurs in the second the attributionaldecision-makingprocess,

Table 4. Regression Analysis of the Suspensions.


(t-Statistics in Parentheses)
Variablea 1 2 3 4
Constant 22.05 81.58*** - 3.20 11.80***
(.65) (7.20) (-.33) (3.29)
Causal Attributions
MRESP 7.43 1.95
(1.26) (1.15)
GRESP 1 L.01* 2.93*
(1.66) (1.84)
FactsoftheCase
CONSEN -8.27 2.18
(-.94) (.80)
DIST -9.96 -1.29
(-1.13) (-.46)
CONSIS 2.65 .28
(.30) (.10)
Moderators
GENDER -18.13** -18.19** 5.89** 6.35
(- 2.08) (- 1.97) (-2.29) (-2.39)
EXP -.55 -.46 .18 .20
(-1.38) (-1.16) (1.52) (1.65)
R-squared .07 .07 .13 .11
F for GENDER Interactionsb .13 .98 1.78 1.04
F for EXP Interactionsb 1.19 1.31 1.56 1.18
Sample Size 133 133 76 76
a For definitions,see note 'a' to Table 1, or text.
b
F-ratiofor interactionswiththe causal attributionsor the factsof the case.
* Significantat the .10 level; ** at the .05 level; *** at the .01 level (two-tailedtest).

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
558 INDUSTRIAL AND LABOR RELATIONS REVIEW

and Models 2 and 4 are estimatesof the expected sign in only one equation (the
reduced-formequation. reduced-formequation for the decision to
In Model 1, GRESP is statisticallysignifi- sustain or deny the grievance); it was
cant with the expected sign, but MRESP is significantwith the opposite sign in one
not significant.EXP and all of the interac- equation (the structuralequation for the
tions are also insignificant.The estimate attributionsof causality to management),
on GENDER, however, is negative and and insignificant in all of the other
statisticallysignificant,indicatingthat the equations, Thus, although the results are
average length of the suspension was 18 mixed, they provide littlesupport for the
days shorterwhen the grievantwas identi- distinctivenessaspect of the attribution
fied as femalethan when the grievantwas model.
identifiedas male. In the equations for the second struc-
In Model 2, none of the factsof the case tural link in the attributionmodel, at least
are statisticallysignificant.EXP and all of one of the variables reflectingthe arbitra-
the interactions are also insignificant. tors' causal attributionswas statistically
GENDER is the only statistically significant significantwiththe expected sign foreach
variable in the equation, indicating,as in of the three decision variables. Thus, with
Model 1, that suspensions for female the exception of distinctiveness,both of
grievantswere 18 days shorter than for the structuralrelations in the attribution
men. model and the reduced-formrelation are
The pattern of results is the same for supported.
the smaller subsample in Models 3 and 4. Moderators
GRESP is significantin Model 3, but none
of the facts of the case are significantin Both of the moderatorswere statistically
Model 4. GENDER is negative and statisti- significant for parts of the attribution
cally significantin both Models 3 and 4, model. The grievant'sgender did not have
indicating that for the subsample of a significantimpact on the arbitrators'
arbitratorswho substituteda suspension attributionsof causality,thatis, in the first
not equal to the delay, female grievants structuralrelation, but it was statistically
received,on average, a suspension 6 days significantin the second structuralequa-
shorterthan male grievants. tion and the reduced-formequation for
two of the decision variables, the award
and the length of suspension substituted
Discussion by the arbitrators. For both of these
The AttributionModel decision variables, the gender effectwas
toward more lenient treatmentof women
The attribution model of arbitrators' than men. This pattern is consistentwith
decision-making in discipline cases was field studies (Bemmels 1988a,b,c and
generally supported by the experimental 1990b; Ponak 1987) and a field experi-
results. The variables reflectingthe ma- ment (Oswald and Caudill 1988) thathave
nipulated facts of the case relating to found more lenient treatmentof women
consensus and consistency were statisti- in arbitrators'decisions, but contradicts
cally significantwith the expected sign in other studies that have found no gender
the structuralequations forthe firstlinkin effects(Block and Stieber 1987; Scott and
the attribution model, and in all the Shadoan 1989) or more lenient treatment
reduced-form equations except the re- of men (Rodgers and Helburn 1985).
gressionfor the lengthof the suspensions In addition to supporting the studies
substitutedby the arbitrators.Thus, the thathave found more lenienttreatmentof
consensus and consistencyaspects of the women, the results provide furtherclues
attributionmodel are clearlysupported. to the source of gender effectsin arbitra-
The distinctivenessmanipulation,how- tion. A shortcomingof the field studies
ever, was not as successful. The distinc- testingwhethergender effectsare present
tiveness variable was significantwith the is that they reveal littleabout the sources

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
ATTRIBUTION THEORY AND DISCIPLINE ARBITRATION 559

of gender effectsin the arbitrators'cogni- tions. None of the factsof the case were
tive decision-makingprocesses. statisticallysignificant. Thus, once the
In the results reported here, all of the arbitrators decided that the grievant
interactionsbetween gender and the facts should be reinstated with a suspension
of the case were insignificant,indicating substitutedfor the discharge, the factsof
that gender does not alter how arbitrators the case had no bearing on the length of
use the facts of the case (such as the suspension chosen. One possible explana-
weights they assign the various facts) to tion for this result is that individual
arrive at their causal attributionsor their arbitrators have a standard suspension
decisions. Gender was also insignificantin that theysubstitutein cases in which they
the firststructuralrelation,indicatingthat reach thisdecision point,but the lengthof
the grievant'sgender did not distortthe the suspension varies among arbitrators.
arbitrators' perceptions of who was re- For example, some arbitratorsmay rou-
sponsible for the incident that led to the tinelysubstituteone-weeksuspensionsand
discharge. others may routinelysubstitutetwo-week
Gender was statistically significant,how- suspensions, regardless of the factsof the
ever, in the second structuralrelationand case. The significantgender effect,how-
the reduced-formrelation,indicatingthat ever, indicatesa double standard,withthe
when the factsof the case and arbitrators' standard suspension being shorter for
perceptions of responsibilityfor the inci- women than men.
dent are held constant, the arbitrators The interactionbetween the grievant's
decided women should receive a more gender and the arbitrators'experience was
lenient penalty than men. This result negativeand statisticallysignificantforthe
suggestsa double standard in selectingthe award decision variable. This negative
appropriate penalty: female gender ap- interactionindicatesthatthe gender effect
pears to be an unacknowledged (and diminishesforarbitratorswithmore expe-
probablyunconscious) but defactomitigat- rience, and reaches zero at 35 years of
ing factorin deciding on the appropriate experience. To the extent that gender
penalty. effects are something to be avoided in
These resultsrelevant to the grievant's arbitrationdecisions, this result suggests
gender are partially consistent and par- that"practicemakes perfect,"although 35
tiallyinconsistentwithprevious studies of years is a lot of practice. Only 30 of the
gender effectsin the attributionaldeci- 230 arbitratorsin the sample (13.0%) had
sion-makingprocess. They are consistent 35 or more years of experience as an
withother findingsthat grievants'gender arbitrator.
affects arbitrators' decisions. They are The arbitrators'experience also had a
inconsistent,however, in terms of where significantmain effect in several of the
gender effectsenter the cognitive deci- equations: more experienced arbitrators
sion-making process. Previous studies were more likely than those with less
have found that gender effectsenter the experience to assign to the grievant the
firststructuralrelation (Deaux and Em- responsibilityfor the incident that led to
swiller 1974; Dobbins 1985; Feldman- the discharge. Similarly,in the reduced-
Summers and Kiesler 1974; Garland and form equation for the decision to sustain
Price 1977; Nieva and Gutek 1980; and or deny the grievance,more experienced
Rose 1978), but not the second structural arbitratorswere more likely to deny the
relation (Heilman and Guzzo 1978). In grievance.These resultsaccord withthose
contrast, these results show that gender of Nelson and Curry (1981),8 but contra-
effectsentered only the second structural
relation.
8 Bemmels (1990a) also found that arbitrators'
In the reduced-formequations for the
decisions are significantlyrelated to their age, but
length of suspensions substitutedby the age may not be a good proxy for experience, since
arbitrators,grievant'sgender was the only arbitratorsbegin practice at differentages, often
significantvariablein the equa-
statistically after retiringfrom a previous career. Nelson and

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
560 INDUSTRIAL AND LABOR RELATIONS REVIEW

dict other studies' findingthat arbitrators' to have been guilty of a lapse of an


experience had no effect on their deci- identicalnature in the past-that is, when
sions (Fleming 1965; Heneman and Sand- there was high "consistency"-the deci-
ver 1983). sions tended to be more unfavorable to
In the second structuralequation for the grievant than when the lapse was a
the award decision variable, however, first offense. The results provide less
more experienced arbitratorswere more support, however, for the predictionthat
likely to award a full reinstatementwith the arbitrators'decisions would be more
full backpay (as opposed to substitutinga favorable to the grievantwhen the griev-
suspension for the discharge) than were ant was said to have performedsatisfacto-
less experienced arbitrators.Thus, it ap- rily on the job in ways unrelated to the
pears that more experienced arbitrators infractionin question-that is, when the
are more likely than less experienced infractionhad high "distinctiveness."
arbitratorseitherto decide entirely for The arbitrators'attributionaldecision-
managementand deny the grievance,orto making process was also found to be
decide entirelyfor the grievantand award affected by two "moderators": the griev-
a reinstatementwith full backpay. Less ant's gender and the arbitrator'syears of
experienced arbitratorsare more likelyto experience as an arbitrator. Those two
choose a middle ground, sustaining the variables, however, entered the decision-
grievancebut substitutinga suspension as making process in different ways: the
a lesser penalty. grievant's gender influenced only the
In terms of attribution theory,.this selection of the appropriate penalty (fe-
pattern suggests that more experienced male grievants were given more lenient
arbitratorsmay have more highly devel- penalties than male grievants), whereas
oped causal schemata than less experi- the arbitrator's experience influenced
enced arbitrators.This interpretationis both the formationof causal attributions
consistent with Kelly's (1972) original and the selection of the appropriate
conception of causal schemata developing penalty (arbitratorswith more experience
with experience. The benefit of many were more likely than less experienced
years of experience may allow arbitrators arbitratorsto judge the grievant,rather
to feel more certain about their causal than the environment,responsible for the
attributionsand, consequently,more com- incident causing the discharge, and more
fortablewithextremedecisions. likelyto eitherdeny the grievanceentirely
or reinstatethe grievantwithfullbackpay,
as opposed to takingthe middle course of
Conclusion reinstatementwitha suspension).
The results of this study generally The finding that these arbitrators,all
support attributiontheoryas an explana- male, tended to impose lesser penalties
tion of how arbitratorsdecide discipline on women than on men is consistentwith
cases. As hypothesized, when the griev- the findings of most other research on
ant's coworkers were said to have some- the subject-both field studies of actual
timescommittedthe same error for which arbitrationcases and experimental stud-
the grievantwas discharged-that is, when ies. Nonetheless, considerable inconsis-
there was high "consensus"-the arbitra- tency exists across the results of those
tors tended to make decisions more studies. As pointed out elsewhere (Bem-
favorable to the grievant than when the mels 1988c, 1991), however, even when
transgressionwas said to be unique to the statisticallysignificant,the gender effects
grievant;and when the grievantwas said are much weaker than the influence of
other factors (such as the nature of the
offense and the grievant's prior record)
Curry(1981), however,also found significantdiffer-
ences by arbitrator'sage that were consistentwith
on arbitrators' decisions. One plausible
their finding of differencesby arbitrator'sexperi- interpretationof the inconsistenciesacross
ence. studies is, therefore,that gender effects

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
ATTRIBUTION THEORY AND DISCIPLINE ARBITRATION 561

are quite weak and are captured at a burn (1985) suggest,arbitrators'decision-


statisticallysignificantlevel only in some, making is somewhatarbitrary!
not in all, studies. This studysuggestssome ideas forfuture
The interpretationof the resultsof this research on discipline arbitration. Since
studyis complicatedby the imperfectfitof grievantgender effectswere found onlyin
the estimated models. Although many of theselectionof theappropriatepenalty,and
the variables are statisticallysignificant, notin thearbitrators'
causal attributions,
fur-
the goodness of fit statisticsindicate that therstudiescould investigatehow and why
most of the variation in the arbitrators' thisdouble standardcomes about. Do arbi-
causal attributionsand decisions is left tratorsbehave paternalistically?Or do they
unexplained. This result is troublesome, practicea formof affirmative actionto com-
since the transcripts received by the pensate for other discriminationagainst
arbitratorswere identical except for the womenin employment? Furtherexperimen-
variables that are controlled in the esti- tal studiescould evaluate whethertraining
mated equations. Arbitratorcharacteristics for arbitrators(similarto that intended to
not controlled for in this study may be reduce certaintypesof ratingerror) might
moderating the decisions, but that expla- reduce or eliminategender effectsin arbi-
nation seems unlikely in light of some tration. Another interestingfinding that
other studies (Heneman and Sandver bears closer examinationis the greaterten-
1983; Bemmels 1990a). Furthermore, dency of arbitratorswithmore experience
other studies (Thornton and Zirkel 1990) to make extremedecisions; research could
have found a similar low level of consis- be done to determinewhethermore expe-
tency and predictabilityin arbitrators' riencedarbitratorshave greaterconfidence
decisions. Perhaps, as Rodgers and Hel- in theircausal attributions.

REFERENCES

Bemmels, Brian. 1988a. "The Effect of Grievants' What Is Skill forthe Male Is Luck forthe Female."
Gender on Arbitrators'Decisions." Industrialand Journalof Personality and Social Psychology,
Vol. 29,
Labor RelationsReview,Vol. 41, No. 2 (January), Pp. 80-85.
pp. 251-62. Dobbins, Gregory H. 1985. "Effectsof Gender on
. 1988b. "Gender Effectsin DisciplineArbitra- Leaders' Responses to Poor Performers: An
tion: Evidence fromBritishColumbia." Academy of AttributionalInterpretation."Academyof Manage-
ManagementJournal,Vol. 31, No. 3 (September), mentJournal, Vol. 28, No. 3 (September), pp.
pp. 699-706. 587-98.
. 1988c. "Gender Effectsin Discharge Arbitra- Feldman-Summers,S., and S. Kiesler. 1974. "Those
tion." Industrialand LaborRelationsReview,Vol. 42, Who Are Number Two Try Harder: The Effectof
No. 1 (October), pp. 63-76. Sex on Attributions of Causality." Journal of
. 1990a. "ArbitratorCharacteristicsand Arbi- Personalityand Social Psychology,Vol. 30, No. 6, pp.
tratorDecisions."JournalofLaborResearch,Vol. 11, 846-55.
No. 2 (Spring), pp. 181-92. Fleming,Robin W. 1965. TheLaborArbitration Process.
. 1990b. "Gender Effectsin GrievanceArbitra- Urbana: Universityof Illinois Press.
tion." IndustrialRelations,Vol. 29, No. 3 (Fall), pp. Garland, H., and K. Price. 1977. "AttitudesToward
513-25. Women in Management and Attributions for
_ 1991. "Research on Gender Effects in Their Success and Failure in a Managerial Posi-
Arbitration."In WilliamKaplan, JefferySack, and tion."JournalofAppliedPsychology, Vol. 62, No. 1,
Morley Gunderson, eds., Labour Arbitration Year- pp. 29-33.
book.Toronto: Butterworths, in press. Green, Stephen G., and Terence R. Mitchell. 1979.
Block, Richard N., and Jack Stieber. 1987. "The "Attribution Processes of Leaders in Leader-
Impact of Attorneysand Arbitratorson Arbitra- Member Interactions."OrganizationalBehaviorand
tion Awards." Industrialand LaborRelationsReview, Human Performance, Vol. 23, pp. 429-58.
Vol. 40, No. 4 (July),pp. 543-55. Greenberg, Martin S., and R. Barry Ruback. 1982.
Campbell, Donald T., and Julian C. Stanley. 1963. Social Psychologyof the CriminalJustice System.
Experimental and Quasi-Experimental Designsfor Re- Monterey,Calif.: Brooks/Cole.
search.Boston: Houghton Mifflin. Heilman, Madeline E., and Richard A. Guzzo. 1978.
Deaux, K., and T. Emswiller.1974. "Explanations of "The Perceived Cause of Work Success as a
Successful Performance on Sex-Linked Tasks: Mediator of Sex Discriminationin Organizations."

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions
562 INDUSTRIAL AND LABOR RELATIONS REVIEW

OrganizationalBehavior and Human Performance, IndustrialRelations,Vol. 20, No. 3 (Fall), pp.


Vol. 21, No. 3 (June), pp. 346-57. 312-17.
Heneman, Herbert G. III, and Marcus H. Sandver. Nieva, Veronica F., and Barbara A. Gutek. 1980.
1983. "Arbitrators'Backgrounds and Behavior.' "Sex Effectson Evaluation." Academyof Manage-
JournalofLaborResearch,Vol. 4, No. 2 (Spring),pp. mentReview,Vol. 5, No. 2, pp. 267-76.
115-24. Oswald, Sharon L., and Steven B. Caudill. 1988.
Kelly,Harold H. 1967. "AttributionTheory in Social "Gender and the Arbitral Decision." Working
Psychology."In D. Levine, ed., NebraskaSymposium Paper, Departmentof Management,Auburn Uni-
on Motivation,Vol. 15. Lincoln: University of versity.
Nebraska Press, pp. 192-238.
Ponak, Allen. 1987. "Discharge Arbitration and
. 1972. "Causal Schemata and the Attribution
Reinstatementin the Provinceof Alberta."Arbitra-
Process." In E. Jones et al., eds. Attribution:
PerceivingtheCauses ofBehavior.Morristown,N.J.: tionJournal,Vol. 42, No. 2 (June), pp. 39-46.
General Learning Press, pp. 1-26. Rodgers, Robert C., and I. B. Helburn. 1985. "The
'Kelly, Harold H., and John L. Michela. 1980. Arbitrarinessof Arbitrators'Decisions." Proceedings
"AttributionTheory and Research." In M. Rosen- of the Thirty-SeventhAnnual Meeting (New York,
zweig and L. Porter, eds., Annual Review of Dec. 28-30). Madison, Wis.: Industrial Relations
Psychology. Palo Alto, Calif.: Annual Reviews Inc., Research Association,pp. 234-41.
pp. 457-501. Rose, Gerald L. 1978. "Sex Effects on Effort
Lord, Robert G., and Jonathan E. Smith. 1983. Attributionsin Managerial Performance Evalua-
"Theoretical, InformationProcessing, and Situa- tion." OrganizationalBehavior and Human Perfor-
tional Factors AffectingAttributionTheory Mod- mance,Vol. 21, No. 3 (June), pp. 367-78.
els of Organizational Behavior." Academyof Man- Scott, Clyde, and Elizabeth Shadoan. 1989. "The
agementReview,Vol. 8, No. 1, pp. 50-60. Effectof Gender on ArbitrationDecisions."Journal
Mitchell, Terence R. 1982. "Attributionsand Ac- ofLaborResearch,Vol. 10, No. 4 (Fall), pp. 429-36.
tions: A Note of Caution." Journalof Management, Thornton,RobertJ.,and PerryA. Zirkel. 1990. "The
Vol. 8, No. 1, pp. 65-74. Consistencyand Predictabilityof Grievance Arbi-
Nelson, Nels E., and Earl M. Curry, Jr. 1981. tration Awards." Industrial and Labor Relations
"ArbitratorCharacteristicsand ArbitralDecisions." Review,Vol. 43, No. 2 (January),pp. 294-307.

This content downloaded from 202.43.95.117 on Sun, 2 Mar 2014 11:15:08 AM


All use subject to JSTOR Terms and Conditions

Anda mungkin juga menyukai