Controling

Controlling Decision-Making Practice in Organizations
Author(s): Kathleen M. Sutcliffe and Gerry McNamara

Source: Organization Science, Vol. 12, No. 4 (Jul. - Aug., 2001), pp. 484-501
Published by: INFORMS
Stable URL: http://www.jstor.org/stable/3085984 .
Accessed: 23/09/2013 20:19
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.
INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Organization Science.
http://www.jstor.org
This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

All use subject to JSTOR Terms and Conditions
Controlling
Decision-Making Practice
in Organizations
KathleenM. Sutcliffe* GerryMcNamara
Departmentof OrganizationalBehaviorand HumanResourceManagement,
ofMichiganBusinessSchool,AnnArbor,Michigan48109-1234
The University
DepartmentofManagement,Broad GraduateSchool ofManagement,MichiganState University,
East Lansing,Michigan48824-1122
umich.edu * mcnama39@pilot.msu.edu
ksutclif@
Abstract and theirassociatedoutcomes.This studyinvestigates the

How are decision practicesfosteredin organizationsand how extentto whichdecisionmakersuse a prescribed decision
are they linked to decision outcomes? This studyaddresses practiceforconsequentialdecisionsand thefactorsthat
thesequestionsby examiningone financialinstitution's efforts influenceits use. We also examinetheoutcomesassoci-
to standardizeand controldecision makingacross geographi- ated with a formallyprescribeddecision-making prac-
cally separatedorganizationalunits.We argue thatdecision- tice-an area thathas received very littleattentionin
makerbehavioris situatedand is not simplya functionof in- decision-makingliterature(Dean and Sharfman1996).
dividual choice. Rather,in organizationalsettingsdecision- Our studyis groundedin two assumptions:first, thatan
makersare subjectto a hierarchyof influencesthataffectthe
understanding of decisionpracticeis fundamental to un-
decisionprocessestheyuse and theirresulting decisionchoices.
To testour ideas, we examined900 borrower"riskrating"de-
derstandingresultingdecisions and decision outcomes,
cisionsand foundgeneralsupportforourhypotheses.Decision and second, thatresearchersstudyingdecision-making
makerswere more likelyto use the prescribedpracticewhen processesneed to pay more systematicattentionto the
decisionswereimportant, whenthedecisiontargetwas known, social contextualforces at work withinorganizational
and when the decision makerwas located in a largersubunit. systems(Mowday and Sutton1993).
Decision makersalteredtheirdecision practicesin the short We investigatethreeinterrelated questions:First,to
term,but in the long termtheyappearedto partiallyrevertto whatextentcan an organizationenacta decisionpractice
theirearlierpractices.Reliance on prescribedpracticefostered and makeit stick?Second,whatfactorsencourageor im-
stabilityin decisions,but surprisinglyappeared to negatively pede theuse of whatis prescribed?Finally,how are pre-
affectfuturejudgments.The findings indicatethatorganizations scribedpracticesand the resultingdecisionslinked?In
can changethemoremicroaspectsofdecisionmaking,butthese
short,thegoal of thisstudyis to shed lighton theability
changes may be transitory. Moreoverthe resultssuggestthat
decision makersmay become complacentwhen theyrelyon
of an organizationto systematizedecision-making prac-
prescribeddecision practices,a tendencythatcan have unto- tice and on the challengesof doing so. These questions
wardconsequencesfortheorganizationsin whichtheyare em- are fundamental to organizationand management theory
bedded. because the designof decision systemsis important for
(Decision-Making
Process;DecisionRoutines;DecisionEffectiveness) organizationaleffectiveness and is an important element
of management-centered conceptionsof organizations
(Dean and Sharfman1996, p. 368; Huberand McDaniel
1986).
To addressthesequestions,we assess theeffectiveness
of one financialinstitution's
plannedefforts to influence
Introduction decisionrationality by formalizing theelementsof infor-
Organizations often go to great lengths to encourage mationthatdecisionmakersare supposedto use formak-
decision-making by the rules. Yet, in spite of the nor- ing important lendingdecisions.Althoughindividualbe-
mative and theoretical emphasis on the use of formal havioris oftenthetargetof plannedchangeefforts, few
decision-making procedures in organizations, surpris- studiesof changeexplicitlyinvestigatechangesin indi-
ingly,littleis actually known about how theyare fostered vidual work behaviors,focusinginstead on summary
ORGANIZATION SCIENCE, ? 2001 INFORMS 1047-7039/O1/1204/0484/$05.OO

Vol. 12, No. 4, July-August
2001, pp. 484-501 1526-5455 electronicISSN

KATHLEEN M. SUTCLIFFE AND GERRY MCNAMARA ControllingDecision-MakingPractice in Organizations
measuresofjob performance orchangesinorganizational makingpracticebyexamining theextentto whichdecision-

outcomes(Robertsonet al. 1993). Thus, this studyhas makersuse specificpieces of information formakinga
potentialto filla gap in theepisodicand plannedchange risk-rating decision.We characterize our study as an in-
(Weick and Quinn 1999). We also examinehow
literature vestigationinto the use of a supplied decision rule-a
decision-makeradherenceto or deviationfroma pre- decisionrule thatis givento a decisionmakerto advise
scribedpracticeis influencedby a hierarchyof factors heror himon how to makea judgment.We use theterm
thatreflectthesociallyembeddednatureoforganizational decision rule ratherloosely, however,to reflecta pre-
decisions (Granovetter1992). Finally,we hope to shed scribedpractice(in thiscase, theuse of particular pieces
lighton the outcomesassociated with the use of pre- of information) ratherthanto reflecta moresophisticated
scribeddecisionpracticesto betterunderstandthe chal- algorithmformakinga judgment.Because the decision
lengesassociatedwiththedesignof decisionprocesses. rule in thiscase concernsthebest pieces of information
to use when makinga particulartypeof judgment,the
practiceis targetedtowardachievingan optimaldecision,
Theoryand Hypotheses reflecting theorganization's underlying valuesandbeliefs
Background thatthepracticewill lead to moreeffective decisionout-
Because decisions can be complex,organizationsoften comes. We use the termsdecisionpractice,prescribed
enactdecision-making proceduresto compensateforthe decisioncriteria,decision rule,andformaldecisioncri-
bounded cognitiveabilitiesof decision makers(March teria interchangeably to reflectthe organizationally pre-
and Simon 1958). It is well establishedthatdecisionmak- scribedinformation thatdecisionmakersare supposedto
ers use a numberof decisional heuristicsor cognitive collectand use in makinga judgment.This prescribedset
shortcuts(Kahnemanet al. 1982), whichcan bias deci- of criteriacan be contrastedwithad hoc criteriathatre-
sion outcomes in systematicways. Moreover,studies flectinformation thatdoes not fitintothe formallypre-
show thatthese heuristicsand shortcutscan systemati- scribedcategories.
cally influenceindividualdecisionprocesses(Northcraft There are two important reasonsto examinehow de-
andNeale 1987) as wellas frameindividual judgments, and cisionmakersgo aboutmakingdecisions.The firstis that
thattheyare elicitedby bothtask and contextualinflu- decision practicesmatter.Studies of decision making
encesinherent in manydecisionsettings(e.g., McNamara in both laboratory(e.g., Arkes et al. 1986, Davis and
and Bromiley1997). Thus, organizationsenactdecision Kottemann1995, Powell 1991) and organizationalset-
rulesto replacetheidiosyncratic heuristicsdecisionmak- tings(e.g., Dean and Sharfman1993, 1996; Eisenhardt
ers may use to make theirjudgmentsand to increasethe and Bourgeois 1988; Fredrickson1984; Fredrickson and
reliabilityof decision makingand the predictability of laquinto 1989) show thatthe way decision makersgo
decisionoutcomesfororganizationsand theirmembers. aboutmakingdecisionsmattersbothto decisionchoices
Standarddecision practicescan enhancetechnicalef- and to outcomes.What emergesfromthese and other
ficiencybecause theychanneldecisionmakers'attention decision-making studiesis thatimportant decisions(i.e.,
and behavior,therebyallowingthemto conservescarce choices) are not simplytheproductof isolatedinforma-
resourcessuch as timeand attention. Yet, organizational tion processing.Rather,different decision procedures
practicesalso providean important stabilizingfunction lead to differentchoices. Moreover,different procedures
(e.g., March and Simon 1958) and are a "centralingre- lead to differentoutcomes.
dientin accountingforthe reliabilityof organizations" A second important reason to pursueresearchin this
(Scott 1995, p. 54). Organizationalpracticesreflectthe domain is to betterunderstandhow decision practices
routineuse oforganizational knowledge(Szulanski1996) (such as the prescribeduse of information) are estab-
and can rendera distinctsource of advantagethatcon- lished,and theoutcomesassociatedwiththeiruse. Some
tributes s effectiveness.
to an organization' Decision prac- evidencesuggeststhatdecisionpracticesare withoutor-
tices increasedecision-making effectivenessby decreas- der, thatinformation and decisions are decoupled,and
ing the uncertaintysurroundingdecision outcomes. thatdecisionmakersprocessinformation withoutregard
Because theyembodyknowledgeof cause-effectrela- to its relevanceto specificdecisions(Cohen et al. 1972,
tionships,operatingpracticesguide decisionmakers'at- Feldmanand March1981,Marchand Olsen 1986,March
tentionto theinformation necessaryto formexpectations 1987). Otherresearchsuggeststhatdecision processes
aboutvariousalternatives (Dean and Sharfman1996) and characterized by thesystematic collectionof information
otherelementscriticalto makinga high-quality decision. relevantto thedecision,and therelianceuponanalysisof
It is on thisaspectthatwe focus. this information in makingthe decision, lead to more
In this studywe investigateorganizationaldecision effectivedecision making(Dean and Sharfman1993,
2001
ORGANIZATIONSCIENCE/Vol. 12, No. 4, July-August 485

Langley 1989). An important unansweredquestionin the In summary,our studybridgeslevels of analysisby

literature
concernswhatfactorsaffecttheextentto which examiningeach decisionand lookingbothat thecogni-
decisionmakerscollectand attendtorelevantinformation tive processes evidenced by the extentto which pre-
when makinga decision. Our studyis focused on an- scribedinformation is cited in supportof each decision
sweringthatquestionby examininginformation handling and at how factorsat each level influencedecision-maker
practices,theireffecton decisionsanddecisionoutcomes, practice.We linkcognitionandcontextbysuggesting that
and how a hierarchyof multilevelinfluencesshapes in- decisionoutcomesaretheresultofthewaydecisionmak-
formation handlingin practice. ers go aboutmakingdecisions.These practicesare influ-
enced by a cascadinghierarchyof influencesrelatedto
The EmbeddednessofDecision-Making
Practice the decision and the contextsin whichdecisionmakers
Our model is derivedtheoretically, based on the simple are situated.
idea thatdecision-maker behavior(practiceand ultimate
choice) in naturalisticsettingsdoes not occurin a social
Antecedentsof Decision Practice
vacuum.Rather,itis contextually embedded(Granovetter
1985, 1992). The idea is thatactionis situatedand can't
be explained solely by referenceto an individual's Decision-LevelFactors.Theorysuggeststhatdecision
choices. Consequently,we propose thatdecision-maker makersare moreor less likelyto use prescribedproce-
behavioris shapedbothby theformalsystemofrulesand dures when certaindecision characteristics are present
organizationalhierarchies,and by the multilevelrela- (e.g., Thompsonand Tuden 1959), yetlittleattention has
tional contexts within which action occurs (Langley been paid to empiricallyexamininghow decisionchar-
1989,O'Reilly and Chatman1996). Formalizationfacili- acteristicsand prescribeddecisionprocessesare linked.
tatesthe diffusionof practicesthroughout an organiza- Experimentalstudiesin this domain have largelybeen
tion.Still,just because decision-making practicesare for- confinedto examininghow a decisionmaker's use ofpre-
malized within an organizationdoes not mean that scribedrulesis influencedby factorsrelatedto thedeci-
decision makerswill use them.For manyreasons,decision task,such as performance feedback(e.g., Davis and
sion makers'operativenormsand behaviorsmay depart Kottemann1995,Powell 1991) or advice abouthow use-
substantiallyfrom formal requirements,and decision ful theruleswill be forparticulardecisions(e.g., Davis
makersmay performa functionin ways thatare at odds and Kottemann1995). Decision rulesare enactedto pre-
withwhatis prescribed.We argue thatadherenceto or ventdecisionmakersfromincorrectly exercisingdiscre-
departurefromformalrequirements is shapedby theim- tion.Presumablythisis essential,especiallywhendeci-
mediatecontextsof action as well as the widerorgani- sions have important consequences.In otherwords,we
zationalcontextin whichdecisionsare made. wouldexpectthatdecisionswithimportant consequences
Applyingthislineofthinking to ourstudysuggeststhat will promptdecisionmakersto be more(or less) inclined
multiplelevels of influencestructuredecision-making to adhere to prescribedpractices.Such consequences
practicewithinorganizations.We considerthreesets of could occur,forexample,whenthe size of thedecision
factors,each operatingat threedifferent levels: thetask/
is large(e.g., involvessignificant
resources)or whenthe
decisionlevel, subunitlevel, and organizationlevel. The
decision concernsan exchangepartner(decision target)
mostimmediatelevel is thatof thedecisioncontext.De-
about whichlittleis known.We considereach of these
cisions withparticularcharacteristics are morelikelyto
be addressedwithparticularmethods(Thompson1967). conditionsbelow.
Consequently,we examine two decision-levelfactors: Theorysuggeststhatdecisionmakerswill use rational
size of thedecisionand durationof therelationship with procedureswhen decisions have high stakes,although
a decision target.At a higherlevel, decision practiceis findingsfromthefew studiesthathave investigated this
influenced by thesubunitcontextin whichdecisionmak- premise empiricallyare mixed. For example, Carter
ersare located,reflected in characteristicssuchas subunit (1971) foundthatdecisionsthatcould seriouslyaffecta
size, subunitspecialization,and past performance. The company'sfutureinvolvedmorerigorousanalysisthan
organizationalcontext,in turn,influencesdecisionprac- less important decisions.Similarly,resultsby Hicksonet
ticethroughstandardization pressures.Factorsat each of al. (1986) suggestthatdecisionmakersare morelikelyto
theselevels play some partin influencing decision-maker use intendedlyrationalprocesses as the size (i.e., re-
behavior,decision choices, and resultingoutcomes.We sources involved) of the decision increases.Dean and
presentreasonsforselectingthesefactorsinthefollowing Sharman(1993) foundno supportforthisrelationship in
section. a studyof strategicdecision making,but attributed the
486 ORGANIZATIONSCIENCE/VO1. 12, No. 4, July-August 2001

resultto constrainedvariancein theirmeasureofdecision makerlacks knowledgeaboutan exchangepartner, he or

importance.Althoughfindingsare mixed,it is plausible she will be particularlysystematicwhen makingdeci-
to thinkthatas thesize of thedecisionincreases,decision sionscrucialto theirfirm'sfutureand will collectand use
makerswill be morelikelyto use prescribedpractices- a wider range of information thatfalls outside of that
presumablybecause adheringto the prescribedprocess whichis organizationally prescribed. in situ-
In contrast,
will resultin moreeffectivedecisionmaking. ationswherea decisionmakerbothknowstheborrower
In additionto instrumental reasons,thereare symbolic andhas developeda trust-based relationshipwiththebor-
reasonsfordecision makersto use prescribedpractices rower,he or she will requirelittlenew information (ad
forhigh-stakesdecisions.Collectingand using the pre- hoc criteria)and will have littleneed to deviatefromthe
scribed informationsymbolizes capable management prescribedcriteriato conducta widerinformation search.
(Dean and Sharfman1993, Langley 1989, Feldmanand The above arguments lead to thefollowinghypothesis.
March 1981). When decisions have importantconse-
with
HYPOTHESIS 2. As thedurationoftherelationship
quences (such as when they involve substantialre-
theexchangepartnerincreases,decisionmakerswilluse
sources),theyare morelikelyto come to theattention of
a greaterproportionofprescribeddecisioncriteria.
superiorsor othertopmanagerswho have theinterests of
thefirmat heart.Consequently, decisionmakersmayfeel SubunitInfluences.Whatdecisionmakersfocuson and
a need to demonstrate capable managementto theirsu- whattheydo dependson the particularmicrocontext in
periors,especially when decisions are important.This whichtheyare locatedbecause attention and behavioris
suggeststhatas thestakesof a decisionincrease,decision situated(Ocasio 1997,p. 190). This meansthata decision
makerswill be morelikelyto relyon prescribedpractices maker'sbehavioris likelyto be influenced by a subunit's
because superiorsare more likelyto perceiveformally existingroutinesthatare tiedto theunderlying systemof
requiredinformation to be a legitimatebasis formaking norms,values, and goals (O'Reilly and Chatman1996).
a decision thaninformation thatlies outsideformalre- As theseidiosyncratic behavioralroutinesare reinforced
quirements.These argumentslead to the followinghy- they become entrenchedand institutionalized (Zucker
pothesis: 1987, p. 456). Thus,it is reasonableto thinkthatauton-
omous unitswill developparticulardecisionpracticesas
HYPOTHESIS 1. As the size of the decision increases,
a consequenceof factorsuniqueto each unit's situational
decision makerswill use a greaterproportionof pre-
context,and theseprocesseswill be reflected in thetype
scribeddecisioncriteria.
of information decisionmakersuse in makingdecisions.
An increasingly prominent featureof transaction-related Under some conditions,thesedifferences can be an ef-
decisions,suchas thosefoundin thecommerciallending fectiveway to achieve differing organizational/subunit
context,concernsthe durationand continuity of the re- goals; underotherconditions,however,thesedifferences
lationshipbetweentheexchangepartners. Researchsug- can impede the success of organization-wide initiatives
geststhatexperiencewithan exchangepartner influences to directjob-relevantbehavior.
action (Granovetter1985). Knowledge,experience,and A numberofcultural,material,and social factorscould
strongexpectationsof trustand abstentionfromoppor- accountfor these idiosyncraticdifferences in decision-
tunism(Granovetter1985, p. 490) accumulateas a rela- makingpractice (Ocasio 1997,p. 190;O'ReillyandChatman
tionshipendures(LevinthalandFichman1988,Seabright 1996) includingsubunitsize, subunitspecialization,and
et al. 1992). Consequently,overtime,exchangerelation- past performance. For example,a subunit'ssize mayre-
ships evolve frominformation exchangesto institution- flecttheextentto whichit emphasizesnormsto comply
alized relationships(Levinthaland Fichman1988). Thus, withformalrequirements. Organizationaltheoryand re-
the durationof an exchange relationshipreflectsboth searchsupportstheidea thatlargerunitsare morelikely
learningand commitment. to requiredecisionmakersto adhereto formaldecision
Taken togetherthese findingshave implicationsfor policies than smallerunits (see Scott 1987). Thus, we
decision-maker behavior.When a decision makerlacks wouldexpectthatdecisionmakersin largerunitswillrely
knowledgeaboutan exchangepartner, he or she willneed more heavilyon prescribedcriteriaand less on ad hoc
a widerrangeof information upon whichto assess pos- criteriathandecisionmakersin smallerunitsthatareless
sible outcomeconsequences.In addition,studiesshow concernedwithformalrulesand procedures.
thatwhenpartiesto a decision are not well known,de- Second,differentsubunitsmayhavetotailortheirprac-
cisionmakersare morevigilantin theirsearchefforts and ticesto perform effectively. It is oftenthecase thatsub-
oftenconsiderissues much more carefullyto preempt unitsdevelopspecificexpertisethatrelatesdirectlyto the
criticism(Tetlock 1992). Consequently,whena decision contextin whichtheyare situated.And researchsupports
2001

the idea thatdecisionmakerstypicallydevelop context- means,theprinciplesof action,interaction, and interpre-

specificheuristicsthatguidetheirdecisionanalyses(Day tationthatguideand constraindecisionmakersin accom-
and Lord 1992, Regehrand Norman 1996). Thus, it is plishingtheirtasks(March and Simon 1958, Thompson
possible thatdecision makerstailortheirdecision prac- 1967). All else being equal, as job definitionsand con-
ticesto accommodateto thespecificdecisionmakingre- creteproceduralspecifications becomemorestandardized
quirements of thecontext.One important aspectof a sub- and formalized,decision makerswill be morelikelyto
unit'scontextis thetypeof customerstheunitserves.In use a prescribedpracticenot only because theyexpect
thecommerciallendingcontext,subunitsoftenspecialize thesepracticesto be of instrumental value, but also be-
and targettheirservicestowardparticularcategoriesof cause thecriteriaare salient,and because adheringto the
customers.It is possiblethatcertainclasses of customers practiceis consideredto be a "legitimate"way to make
may be morereadilyevaluatedusing the generallypre- a decision.
scribedcriteria,while othersmay requireotherad hoc Thereare also reasonsto expectthatefforts to formal-
criteriato be effectively evaluated.The idea is thatthe ize decision-making practicewill be modestand/ortran-
routinesthat develop aroundborrowerevaluationsare sient.First,althoughformalcontrolsystemsaregrounded
driven,in part,by thecharacteristics ofthecustomerbase in theidea thatpeople will unfailingly acceptlegitimate
thatthe branchserves.Supportforthisline of thinking authority, psychologicaltheoriesof reactancehighlight
comes fromresearchon service firms(Nayyar 1990, thatpeople have strongdesiresto maintaintheirpersonal
Brushand Artz1999) suggestingthatthecriticalresource control and freedomof action (Worchel and Brehm
bases of service firmsare intangible.Consequently,in 1971). Second,new behaviorscan also be difficult to es-
contrastto moretraditional manufacturing industries,ser- tablishbecause insufficienttimeis spenton "unfreezing"
vice firmsmaynotbe as effectively evaluatedusinga set and cognitiverestructuring (Schein 1996). Moreover,al-
of financialratios (of which the prescribedcriteriaare thoughit is possible thatdecision makerswill increase
composed). Thus, branchesthatfocus on servicefirms theiruse of standarddecision practicessoon aftera
mayrelyon ad hoc criteriamorethanbranchesthatfocus changeis enacted,theymaysubsequently revertto earlier
on manufacturing firms. learnedbehaviors(Weick 1996) or be less motivatedto
Finally,a subunit'spast performance could also ma- rely on the new proceduresonce standardization pres-
teriallyinfluencedecision-makerbehavior.Theoriesof surestaperoff(Tyreand Orlikowski1994). While these
success suggestthatdecisionmakersin highlyperform- argumentscall intoquestionthe amplitudeand sustain-
ing unitsare more likelyto be deemed competentand abilityof formalizationefforts, theydo not negate the
may be subjectto less scrutinythandecision makersin generalpremisethattheseefforts will have some effect.
poorlyperforming units(e.g., Miller 1993, Sitkin1992). Consequently,we hypothesizethefollowing:
In contrast,decision makersin poorlyperforming units HYPOTHESIS 4. As organizationalpressuresto stan-
may feel added pressureto conformto the rules forin- dardizedecisionmakingbecomestronger, decisionmak-
strumental reasons (because the practicesare presumed ers will use a greaterproportionofprescribeddecision
to lead to moreeffectivedecisions,as notedinHypothesis criteria.
1) or forsymbolicreasons,to demonstrate legitimateand
capable behaviorto superiorsor otherswithwhom the Outcomes of a Prescribed Decision Practice
decisionmakermayconsultor who maymonitorthede- Organizations typicallyseekto standardizedecisionmak-
cision maker'sbehavior.This leads to thefollowinghy- ing forseveralreasons,includinga desireto increasethe
pothesis: consistencyof the decision-making processbothwithin
and across organizationalunits,as well as to raise the
HYPOTHESIS 3. Subunitcharacteristics (size, customer consistencyof thedecisionevaluationsover time.Deci-
base, past performance)will be relatedto thedegree to sion rules are likelyto increasethe stabilityof decision
whichtheunit'sdecisionmakersuse prescribeddecision assessmentsovertimeas theyalterboththerangeof in-
criteria. formation consideredas well as thecognitiveprocesses
used toevaluatetheinformation. Fundamentally, decision
Organization-LevelContext.Decision-makerbehavior rules are designedto codifyorganizationallearningand
is regulatedalso by the wider organizationalcontext, constraintheattention of thedecisionmakersso thatthey
whichis substantively shapedbythelargerorganization's ignoreextraneousinformation thatsimplyadds noise to
rules and proceduralspecifications.Organization-wide theirdecision evaluationprocesses (Levitt and March
rules,in particular,
signalthevalue and legitimacyof ap- 1988). Alternatively,decisionrulesmaycause a decision
propriatebehaviorsand therebyprovide,throughformal makerto ignore,overlook,or otherwisebecomeawareof
488 ORGANIZATIONSCIENCE/Vol. 12, No. 4, July-August 2001

criticalfactorsthatindicateimportant changesin thede-in the studyhad been loosely affiliatedwiththe parent

cision environment.In eithercase, decision rule use bank for more than 10 years beforethe changes were
shouldengenderstabilityin assessments. initiated.
Researchsuggeststhattheamountand typesof infor- Priorto the year of the reorganization, the bank was
mationused in assessing decision targetsare relatedtoorganizedas a holdingcompany,witheach brancha sepa-
thelevel of certaintythedecisionmakerexperienceswith rateentity.Then,in responseto a regulatory change,the
bank changedbothits legal structure
thedecision(Leblebici and Salancik 1981). For example, (fromseparateaf-
filiated banks to an integratedbank with multiple
in a studyof loan decisionmaking,Leblebiciand Salancik
(1981) foundthatdecisionmakerswho perceiveda high branches)and management practices.One setof changes
degreeofcertainty abouta decisiontarget(loan applicant)
was focusedon standardizing riskassessmentandlending
were less likelyto gatherand considerad hoc informa- policies and the associated decision-makingpractice
tion-informationnot partof thatnormally/formally across branches.It is importantto note that although
re-
quired.In otherwords,lendersaremorelikelytoperceive manyorganizational changeswereenacted,to ourknowl-
a highdegreeof certainty abouta decisiontarget,suchas
edge, no significantchangesweremade to theformalre-
a borrower,when theyperceivethe qualityof the bor- wardsystemduringtheperiodof thestudy.
rowerto be steady.Takentogether, thesefindings suggestAlthoughorganizationsoftenintroduceformalproce-
thatthe use of prescribedrules may signal a decision duresto increasedecision-making efficiency, interviews
maker'slevel of certainty abouta borrowerand thelike-withseniormanagersrevealedthatincreasedefficiency
lihoodthattheywill stickwiththeirpreviousassessment perse was notthegoal of thiseffort. Risk-rating decision
of theborrower'ssituation. practiceswere alreadyroutinizedthroughout theorgani-
In additionto influencing thetypeof information zation,but decisionpracticeswere idiosyncratic
de- to each
cision makerstake into account,standardizeddecision organizationalsubunit.Top managementwantedto in-
processeswill also influencethe way in whichdecision crease both the predictability of decisions and decision
makersprocessinformation whenevaluatingdecisiontar-
effectiveness by codifyingorganizationallearning.Spe-
gets.Researchersdiscussingtherole of automaticversus cifically,interviewswithseniormanagersrevealedthat
activecognitiveprocessingand itslinkwithdecisionpro- the institution was interestedin (a) developinga "best
cedures(e.g., Dutton1993, Gioia and Poole 1984) sug- practice,"(b) developing a consistentprocess across
.gest thatdecision makersusing standardizedrules will banks (i.e., subunits),and (c) ensuringthatthe process
engage less in activeprocessingand morein "scripted" was uniformly understoodand implemented. As manag-
automatic processing. Decision makers in automatic ers pointedout,thestandardization effortlengthened the
reviewprocessand resultedin longerloan reviewdocu-
mode are less likelyto see theneed fora changein their
decisionassessmentthandecisionmakersin activemode. mentsand moresystematic reviewsby superiors.
Thus,overtime,we would expectto see evidenceof sta- The year beforethe new decision practiceswere im-
ble patternsin theirassessments.For thisreasonand theplemented,the bank convenedan internalpanel of top
reasonsincludedin thepreviousparagraphwe hypothe- managersand seniorlendersand used a delphiprocessto
size: create the set of "standardcriteria"that loan officers
HYPOTHESIS 5. The greater the proportionof pre-
would be asked to use. To generatethe set of criteria,
panel membersdrew fromtheirrelevantcollectiveex-
scribedcriteriadecisionmakersuse in time(t), theless
pertise, an assessmentof currentindustry practices,and
likelytheywill be to change theirdecision assessment
information gleaned from pertinent professional publi-
fromtime(t - 1) to time(t).
cations.The finalset of criteriacomprisedeightcatego-
ries of factorsdeemedby bank managersto be themost
Method importantcriteriafor assessing the risk of commercial
Research Context borrowers.The criteriaincluded conventionalfactors
To studyan organization'sabilityto systematizeday-to- commonlyused in the bankingindustryto evaluate the
day decision practiceswe examineda contextin which healthof an organization,includingcash flow,finance-
decisionmakersmakingimportant decisionswere faced able assets, net income, liquidity,leverage,collateral,
with a planned change in decision-making procedures. firmsize, and sourcesof loan repayment.
We studiedthe decisionsmade by loan officerslocated The bank enacted the refineddecision-makingprac-
in fivebranchesof a largecommercialbank in theMid- ticesin thefollowingyearand trainedall lendingofficers
west thatunderwenta structural reorganizationand re- in usingtheprescribedset of risk-rating decisioncriteria
designin management procedures.All branchesincluded to assess the riskof commercial borrowers. The formal
ORGANIZATIONSCIENCE/Vol. 12, No. 4, July-August 2001 489

criteriawere printedon a table titledthe "Risk-Rating ways. Commerciallendingdecisions involve allocation

Implementation Guide," whichwas includedin a newly offirmresourcesforexpectedgainwherethereis a degree
developedcommerciallendingproceduremanual. of uncertaintyaboutthedecisions' returns and risks,and
In thisstudy,we definethecriteriaincludedon therisk- wherethesuccess of thedecisionis onlypartlyunderthe
ratingguide as "prescribeddecisioncriteria"because in controlofthedecisionmaker.Additionally, as withmany
its trainingmaterials,thebankidentified thesecriteriaas organizationaldecisionprocesses,one employeeleads on
beingtheprimaryfactorsthatloan officersshoulduse to the decision, but supervisorsregularlyreview and ap-
assess borrowerrisk.The bank did not go so faras to provethedecisionmade by thelendingofficer.On par-
identify thesecriteriaeitheras a minimumor maximum ticularlydifficultor importantdecisions,a group may
set of decision criteriato be used in makinga decision. work on the decision. Not surprisingly, the degree of
Rather,top managersidentified thesefactorsas themost managerialoversightdirectlyrelates to the monetary
important evaluativecriteria.In reviewingborrower files, value of the decision. This structured decision process
however,we foundthatdecision makersdid not simply withsignificant humanjudgmentis verysimilarto the
rely on the recommendedcriteria.Bankers oftencited typesof processesfoundin case studiesoforganizational
othercriteriaas thebasis fortheirborrowerassessments. decisionprocesses(e.g., Bromiley1987). Andthesechar-
In contrastto the formallyprescribeddecision criteria, acteristics(individuallead, managerialoversight,and
we definethesecriteriaas ad hoc because theywerenot groupdecisionson majorfinancialoutlays)can be found
includedin reference materials, werenotdiscussedduring in manycorporatedecisionsincludingproductdevelop-
theformaltrainingof commerciallenders,and were not ment,marketexpansion,capital investment, and hiring.
includedon therisk-rating guidein thelendingprocedure In short,whileourstudyof a singletypeof decisionmay
manual. notgeneralizeto all businessdecisions,commerciallend-
It is importantto notethattheprescribeddecisioncri- ingconstitutes an importantkindofbusinessdecisionand
teriaare primarilyfocusedon information relatedto the incorporates process,context,and contentthataresimilar
financialstrength of thecompany(i.e., cash flow,netin- to a rangeof businessdecisions.
come,liquidity,leverage,financeableassets),whereasthe
ad hoc criteriaare focusedon information relatedto man- Sampleand Data
agementcapabilities(e.g., management succession,qual- We collectedtheinterviewand archivaldata to testour
ityofmanagement, experiencewiththebusiness),market hypotheses fromthecommunity bankingdivisionofeach
positionof the firm(e.g., firmefficiency, age of corpo- ofthefivebranchesspecializingin lendingtocommercial
ration,marketposition,reputation, rateof growth),and borrowers.Althoughthere were more than 30 bank
thecharacteristics of thefirm'senvironment (e.g., indus- branchesin themetropolitan area, mostbranchesserved
try,productdemand,cyclicalityof industry). privatecustomers,and the commercialloan operations
Althoughthiswas a convenientsample to investigate wereconcentrated in fivebranches.
how decisionpracticesare fosteredin organizationsand Because thebanktypicallymaintainsongoingrelation-
how are theylinkedto decision outcomes,we chose to shipswithcommercialcustomers, theresponsiblelending
studydecisionpracticesin thissettingfortwo otherrea- officeris requiredto conductan annualreviewof each
sons.First,thiscontext(i.e., commerciallendingdecision borrower.The purposeof theannualborrowerreviewis
making)enabledus to collectdata on a largenumberof to reevaluatethe currentcreditworthinessof a borrower
similardecisionsmade over an extendedperiodof time. using a seven-pointratingscale, with a one being the
Althoughmanycorporatedecisionsinvolve substantive lowestriskand a sevenbeingthehighestrisk.
evaluationof risk,in many instances,especially when The risk-ratingdecisionis themostcriticaldecisionin
decisions are infrequent and unique,the evaluationsare theborrower reviewprocessbecauseitlargelydetermines
notrecordedand information used to make thedecision whethera new loan is approved,whetheran existingre-
is not systematically retained.The commerciallending volvingloan is continued,the interestrateschargedon
contextovercomestheseproblemsand allowed us to ex- theloan, theloan-lossreservesthatthebankwill hold to
amine decision characteristics, the information used to accountforthepossibilityof default,thelevel ofongoing
make the decision, and the decision outcome across a monitoring of the borrowerand the loan, and the unit
rangeofverysimilarjudgmentaldecisionswithina single withinthebankthatwill be responsibleforthemanage-
organizationalsetting. mentof theborrowerrelationship. Thus,it is thecentral
Second, althoughwe cannotdirectlyassess thegener- decision in a lendingrelationshipwitha borrower.Al-
alizabilityof thefindings, commerciallendingdecisions thoughthe specificsof the borrowerevaluationprocess
parallel a range of riskybusiness decisions in several varyfromloan to loan,to assess theriskof theborrower,
490 ORGANIZATIONSCIENCE/VOl. 12, No. 4, July-August 2001

the lendingofficertypicallyreviewsthe borrower'sfi- coded each of the pieces of information cited by loan
nancial statements, meets withthe managementof the officers foreach risk-rating decisionintoone oftwocate-
firm,and toursthebusiness'sfacilities. gories:prescribedor ad hoc. A piece of information was
Once the lendingofficercompletesthe evaluationof categorizedas "prescribed"if it fellintoone of theeight
theborrower,he or she preparesa CreditApprovalPre- categoriespublishedon the risk-rating guide or as "ad
sentation(CAP). The CAP is the writtendocumentthat hoc" if the criteriondid not fitintoone of the standard
supportsthe borrowerevaluationand approval.As part categories.The interrater reliabilityusingCohen's Kappa
of theCAP, thelendingofficerreportsthecriteriathathe (Cohen 1960) was k = 0.7126, p < 0.0001. The mag-
or she utilizedin assessingtherisklevel of theborrower. nitudeof theteststatistic(z = 13.29) suggeststhatthere
A sample of each branch'sloan filesare auditedby an was a substantiallevel of agreementbetween coders
internalauditinggroupon an annual basis, and the au- (Landis and Koch 1977). In thefewinstanceswherecod-
ditorsroutinely questionwhether thelendingofficerprop- ers disagreed,the disagreements were resolvedthrough
erlyassessed theriskof theborrowers.Thus,theCAP is discussion.
consideredto be a veryimportant document. To derive our primarydependentvariable, we first
We collected annual borrowerreview data for 223 countedthenumberof prescribedcriteriacitedforeach
firmsfora periodof six yearsfromthe loan filesat the decisionand thencountedthenumberof ad hoc criteria
fivebankbranchesstudied.These 223 firmsrepresent allcitedforeach decision.We thencalculatedthemainde-
of thesmallbusinessborrowers(less than$20 millionin pendentvariableby dividingthe numberof prescribed
sales) at these fivebrancheswithloan sizes of at least criteriaby thetotalnumberof criteriacitedfora decision
$100,000, the minimumloan size requiringfull loan and multiplying by 100. The othertwo measureswere
documentation. We collecteddata thatincludedthedate simplecounts.The percentageand absolutemeasuresof
thatthelendingrelationship was established,theamount decision rule use distinguishbetween "degree" and
of theloan, theriskrating,and thepieces of information "amount."We transformed thedependentvariablesto ad-
(i.e., decisioncriteria)lenderscitedas thebasis fortheirdress issues of nonnormality in theirdistribution. We
risk-rating evaluation.The unitof analysisin thisstudy transformed thepercentageof prescribedcriteriaby sub-
is the risk-rating decision.Because of multipleyearsof tractingthe percentagevalue froma constantand then
data foreach firm,the data set included900 complete takingthelog of thatvalue. We transformed thenumber
observations.Because some borrowersenteredthebank of prescribedcriteriavariableby subtracting it froma
duringtheperiodof thestudyand othersleft,we do not constantand thentakingthe square root of this score.
have a fullsix yearsof data forall borrowers. Finally,we transformed thenumberof ad hoc criteriaby
addingone to its value and thentakingthe log of that
Measures value. In presenting ourresults,we reversescoretheval-
Dependent Variables. We created three dependent ues for the percentage and numberof prescribedcriteria
measuresforthefirstanalysesexaminingtheantecedents to preserve the low to high ordering.
of decisionpractice.The percentageof thetotalnumber The outcome associated with theuse of theprescribed
decisioncriteriacitedforeach decisionthat
of risk-rating criteria in which we were primarily interestedwas the
reflectedone oftheorganizationally prescribedcategories stability in therisk-rating assessment. To assess whether
constitutestheprimary dependentvariable.We notedear- or not there was evidence of stability,we measuredthe
lierthattheprescribedcriteriareflectthemostimportant risk assessment of the borrower in both theprior(t-]) and
factorsthatloan officersshould use to assess borrower the current year (t). We then created an indicatorvariable
risk,buttheyneitherconstituted a minimumnora max- to measure whether the risk assessment of the borrower
imumset of decisioncriteriato be used in makinga de- had remainedthesame or changed,indicatingstability or
cision. Thus, decision makerscould includead hoc cri- instability in the decision assessment.
teriato supporttheirdecisions.Our primarydependent IndependentVariables. We measuredcurrentsize of
measureprovidesan assessmentof the extentto which theloan in millionsof dollars.We assessed theduration
lendersused moreor less of theseprescribedcriteriarela- of therelationship betweenthebankand theborrowerin
tiveto thetotalnumbercriteriatheyused.We also created termsof thenumberof years.Data forbothof thesevari-
a measureof the absolutenumberof prescribedcriteria ables wereobtainedfromtheloan reviewdocuments.
cited for each decision and a measureof the absolute We assessed threebranchvariables-branchsize, unit
numberof ad hoc criteriacited for each decision. expertisein termsof customerbase, and priorperfor-
To createthe dependentmeasures, three coders,two mance-to assess theextentto whichsubunitcharacter-
researchers and one seniorbank executive independently isticsare associatedwithuse of decisioncriteria.The size
ORGANIZATIONSCIENCE/Vol. 12, No. 4, July-August 2001 491

of thebankbranchwas measuredusingtheaveragesize "simultaneously investigate relationshipswithina partic-

ofthecommercialloan portfoliowithinthebranchin mil- ular hierarchicallevel, as well as relationshipsbetween
lions of dollars. We assessed the customerbase of the or across hierarchicallevels" (Hofmann1997, p. 726).
bank by calculatingthepercentageof thebranch'scom- HLM models bothLevel 1 randomeffectsand Level 2
mercialcustomersthatwere frommanufacturing and higherfixedeffectssimultaneously
as op- usinga multilevel
posed to serviceindustries.Each branch'spriorperfor- regression-based approach(Hofmann1997,p. 729-730).
mance was measured as the net income for lending There is one significant limitationin the HLM method
dividedby thetotalsize of thebranch'slendingportfolio. thatpreventsus fromusingitin thisstudy.HLM requires
These threemeasureswerecalculatedon an annualbasis. the data to be fullynestedin one hierarchicaldirection.
We assessed the strength of theinstitution's
efforts Ourdataarecross-nested.
to Ourdataarenestedbothwithin
standardizedecision practiceacross the organization's borrowers acrossyearsas well as withinbranchesbyyear
geographicallyseparated subunitsusing a perceptual and finallyby levelof standardization (whichis measured
measurebecause we wereunableto objectivelymeasure on an annualbasis). While we are nothypothesizing ef-
standardization. As noted,the bank undertooka major fectswithinborrowersacross years,we need to control
standardization effortduringthe timeof the study.We forthissince,withthepanel natureof thedata,thereis
asked threeseniormanagerswho had been in top-level likelyto be significant autocorrelationacross the years
managementroles throughout the studyto completea fora givenborrower.Althoughthe mostrecentversion
shortquestionnaire to assess thedegreeof standardization
of theHLM software(HLM 5) allows fortheestimations
by yearacross theentirebank.Our measureof standard- of autocorrelation in thedata,thisis onlytrueif theau-
izationconsistedof fourquestionsratedon a seven-point tocorrelation occursin thesame nestingdirectionas the
Likert-type scale. ThreeitemswereadaptedfromVan de Level 1 variablesincludedin the analysis(i.e., autocor-
Ven and Ferry's(1980, p. 161-162) measureof unitstan- relationcan onlybe estimatedatLevel 1). In otherwords,
dardization.A sample itemincluded"How manyrules with the structureof our data and hypotheses,HLM
and procedureswerethereforconductingloan reviewsin would allow forautocorrelation acrossdecisionswithina
thefollowingyears?"The fourth questionaskedtopman- branchfora givenyear,butnotwithinborrowersacross
agersto assess thedegreeof standardization across sub-
years.
units(branches).The Cronbach's alpha across the four Repeatedmeasuresregressionaddressesthemultilevel
itemswas greaterthan0.98 foreach informant. We alsoissue using generallinearmodel principles(Cohen and
calculatedtheintraclasscorrelationto testforinter-rater Cohen 1983). Withrepeatedmeasuresregression, there-
reliabilityand foundthattherewas significant agreementsearcherfirstpartitionsthevarianceinthedependentvari-
amongthecoders(r1= 0.51) (James,1982). We averaged able to reflectthevariationin thecriterion thatoccursat
the responsesacross questionsand informants to get an
each of the levels thatis being modeled.The researcher
overall indicationof the strength of standardization thenuses resultsfroma hierarchical
by regressionanalysis,
year.As we expected,thedata indicatedthatthestrength starting withvariablesat thelowestlevel of analysis,to
of standardization increasedover the timeperiodof the testforthedegreeto whichthevariablesexplainthevari-
study. ation in the dependentvariablethatoccurs at the same
hierarchicallevel as the currentindependentvariable.
Analyses RMR also requiresthattheresearcher employtheappro-
We testedHypotheses1-4 using repeatedmeasuresre- priatedegreesoffreedomto calculatesignificance ateach
gression.As withmuchorganizational research,thisstudy level of analysis(Cohen and Cohen 1983,Hollenbecket
includes variables from multiple levels of analysis al. 1993). As we describebelow,RMR addressesthevi-
(i.e., decision-level,branch-level,and organizational- olationsthatoccur when researchersanalyze multilevel
level variables), necessitatingthat we use an analytic data usingOLS regression.
technique that appropriatelymodels nested data. We First,the methodallows researchersto examine the
identifiedtwo techniquesthathave been used to address abilityofa givenindependent variabletoexplainthevari-
the issue of multilevelanalyses in organizationalre- ation in the dependentvariablethatoccurs at the same
search: hierarchicallinear modeling(HLM) (Bryk and level as theindependent variablewhileholdingthevari-
Raudenbush 1992) and repeated measures regression ance at theotherlevel constant.That is, whenanalyzing
(RMR) (Cohen and Cohen 1983). HLM has receivedin- relationships at the individuallevel, the effectfororga-
creased attentionin the organizationalfieldin the last nizationis controlled(creatingindependenceamongthe
several years as being a flexiblestatisticalmethodfor lowerlevel scoresthatare now residuals),and whenex-
examiningnesteddata because it allows a researcherto aminingeffectsat the organizationallevel, lower level

responsesareaggregatedto arriveat a single,independent 5 is dichotomous.The independent variablein whichwe

scoreforeach organization. were interestedwas thepercentageof prescribedcriteria
Second,theF-testforeach variableis calculatedusing cited. To controlforextraneousmacroeconomiceffects
the appropriatedegrees of freedom.For example,in a thatcould influencetheoverallqualityof the loan port-
studywith400 observationsthatare nestedwithin40 or- folio,we includeda controlvariableto accountforthe
ganizationalunits,a variablethatvaries at the decision macroeconomicconditionsof theregionin whichall of
level would be examinedusing 400 less the numberof thebankbrancheswerelocated.Using data fromtheSta-
degreesof freedomforthevariablesin themodel,or 359 tisticalAbstractsof theUnitedStates,we measuredthe
degreesof freedomifthestudyincludedone independent economichealthof thebank's metropolitan regionas the
variableat each level. However,variablesthatresideat percentchange in per capita net income over the time
the organizationallevel would be testedusingtheorga- periodduringwhichwe examineddecisioneffectiveness.
nizationalunitnumberof degreesof freedom,or 38 de-
grees of freedomif the studyincludedone independent Preliminary Analysis.Our theorizing, in part,is based
variableat each level. Because the individualresponses on the assumptionthat the criteriathatdecision makers
are effectivelysummedto createa singlescore foreach reportedreflectthecriteria they actually used in making
organization,these are independentobservations.This a decision.We cannotdirectlyassess thedegreeto which
methodhas been widelyemployedin themedicallitera- information reportedin theCAP matchestheactual cri-
ture(e.g., Carteret al. 1986, Jacobset al. 1999, Balanda teriaused because thereis no way to systematically cap-
et al. 1999) and has been increasinglyused to examine tureproxyvariablesto measurethead hoc criteria.Still,
multilevelmodels in the organizationalliterature(e.g., we attempted to assess thisassumptionindirectly by ex-
Hollenbecket al. 1995, Hollenbecket al. 1998, Ryan et aminingthe extentto whichproxiesforthe prescribed
al. 1999, Walkerand Smither1999, Hedlundet al. 1998, criteriapredicttheriskratingassignedto a borrower.In
Phillips2000). Given thattherepeatedmeasuresregres- our data set we were able to identifyfinancialvariable
sion methodallows us to appropriately testa multilevel proxiesforsix of theeightprescribedcriteria:cash flow,
model while also allowingus to controlforserialcorre- profitability,liquidity,leverage,collateralmargin,and
lationin the data, we chose to use thisanalyticmethod size. We split a sample of 540 observations,all obser-
to studythehypothesized relationships.' vationsforwhich we had the necessaryfinancialdata,
We enteredtheindependent variablesintotheanalyses intotwoequal sets:thoseobservationsforwhichtheloan
in threephases because theyreflectthreedistinctlevels
officerhad citedan above averagenumberof prescribed
of aggregation.We firstenteredsize of the loan and re-
criteriaand thoseforwhichthe loan officerhad citeda
lationshiptenurebecause theyare measuredat thelowest
below average numberof prescribedcriteria.We used
level of aggregation:theindividualborrowerdecision.In
logisticregressionto examinetheextentto whichthesix
the second series of steps,we added variablesthatare
variablespredictedthe risk ratingassigned to the bor-
measuredat thebranchlevel: subunitsize,customerbase,
rower.The resultsprovidesupportfortheassumption that
and priorunitperformance. In the finalstep,we added
thecriteriareportedby decisionmakersreflectthedeci-
the organization-levelindependentvariable: degree of
sion criteriatheysubstantively used in theirevaluations.
standardization.This entryprocess allowed us to focus
on the abilityof the variablesto explainthevariancein First,we foundthattheproxiesfortheprescribedcriteria
have greaterexplanatory powerin predicting theriskrat-
decisionpracticethatoccurredat theirrespectivelevel of
analysisas well as to use theappropriate degreesof free- ingof theobservationswhereloan officers citeda greater
dom in assessingthe significanceof the contribution of (above average) numberof prescribedcriteriathanfor
each variable.We used thegeneralizedleast squaresre- observationswhereloan officerscitedfewer(below av-
gression module in SAS to generatethe information erage) (X2 = 216.20 vs. x2 = 107.53,respectively). Sec-
needed at each step to calculate the repeatedmeasures ond, a greaternumberof the individualvariableswere
regressionstatistics,and the Yule-Walkerprocedureto significantfortheabove averageobservations thanforthe
correctforfirst-order autocorrelation.We also corrected below averageobservations.For theabove averagesam-
for higherorder heteroscedasticity using the GARCH ple, theproxiesforprofitability, liquidity,leverage,size,
procedurein SAS. The resultingstatisticsindicatedthat andcollateralmarginwereall significant ateitherthe0.01
theseprocedureseffectively correctedforautocorrelation or 0.05 levels. In contrast,forthebelow averagesample
and heteroscedasticity. onlyprofitability and leveragewere significant at either
We testedHypothesis5 using logisticregressionbe- the 0.01 or 0.05 levels. These resultssuggestedthatthe
cause thedependentvariableassociatedwithHypothesis decision criteriareportedin the CAP reflectthe actual
2001

information
decision makersused in makingtheirjudg- We includedthreebranchcharacteristics to examine
mentsand allowed us to move ahead to testthehypoth- the possibilitythatsubunitfactorsdrivethe differences
eses. in decision practice(Hypothesis3). Only subunitsize
demonstrated a consistentsignificant effect.Consistent
withour reasoning,lendersin largerbranchesdemon-
Results stratedgreaterrelianceon prescribedcriteriain theirde-
Table 1 presentsa year-by-year distribution of thedeci-
cision analyses (p = 0.05). Interestingly, the effectof
sion criteriacited by decision makers.Decision makers
bureaucracyseemsto cause decisionmakersto use fewer
used a combinationof prescribedand ad hoc criteriain
ad hoc criteria(p < 0.05) as opposedto a greaternumber
makingtheirrisk-rating decisions,althoughtheyused a
of prescribedcriteria.The customerbase of the branch
higherpercentageof prescribedcriteria(85.7%). This is
and priorunitperformance appearedto have little,ifany,
not surprisingbecause these criteriaare formallyre-
effecton the decision criteriareportedby commercial
quired,legitimatelyaccepted,and presumedto be fun-
lenders.Thus,we findsupportforHypothesis3, butonly
damentalforassessingthe financialhealthof a corpora-
forsubunitsize.
tion. Still, on average,decision makerscite only about
Hypothesis4 predictsthatthe strongerthe organiza-
halfof thetotalnumberof prescribedcriteriacategories
tion's effortsto standardizedecision practice,the more
in theirdecisionanalyses.
decisionmakerswill use prescribedcriteria.The signsof
Table 2 reportsdescriptivestatisticsand correlations
the parametersare in the hypothesizeddirectionforall
betweeneach of the variablesincludedin the analyses.
threedependentvariables,but theyare not significant.
The resultsshowa negativecorrelation betweenthenum-
Thus,theresultsdo notsupportourhypothesis. However,
ber of prescribedand thenumberof ad hoc criteria.The
thisis notsurprising because thetesteffectivelyincludes
resultsof the repeatedmeasuresregressionanalysesfor
only six observations,one foreach year of data in the
our primarydependentvariablesare shownin Table 3.
sample.Hence, our testshave littlestatisticalpower.To
Hypothesis1 statesthatas the size of thedecisionin-
further examinetherelationship betweenperceivedstan-
creases,decisionmakerswill use a greaterproportion of
dardizationand decision rule use, we used ANOVA to
prescribedcriteria.As shownin Table 3, thepercentage
examinethe extentto whichrelianceon prescribedcri-
of prescribedcriteriacited is positivelyrelatedto loan
teriavariedovertheyearsofthestudy.We foundsystem-
size (p < 0.01), supporting Hypothesis1. Also in support
atic differences across the years(F = 3.92, p < 0.01).
of Hypothesis1, we founda positiveassociationbetween
We used a Bonferroni testto examinethedifferences and
the numberof prescribedcriteriacited and loan size (p
foundthatwhile the relianceon prescribedcriteriawas
< 0.01) and a negativerelationship betweenad hoc cri-
higherin 1988, the year thatthe decision process was
teriaand loan size (p < 0.01).
standardized, thanin either1986 or 1987 (t = 3.71 and
Hypothesis2 statesthatas thedurationof therelation-
t = 3.80, p < 0.01), the relianceon thesecriteriawas
ship withthe borrowerincreases,decision makerswill
significantlyless in 1991 thanin 1988 (t = 2.76, p <
use a greaterproportion of prescribedcriteria.As shown
0.05). The patterns aresimilarfortheothertwodependent
in Table 3, the longerthe relationship,the greaterthe
measures.This suggeststhatthestandardization effort in-
percentageof criteriacitedthatareprescribed(p < 0.01).
fluenceddecisionmakers'practicesin theperiodimme-
Moreover, relationshipdurationis significantly nega-
diatelyfollowingtheintroduction ofthenewpractice,but
tivelyassociatedwiththenumberof ad hoc criteriacited
theeffectsappearedto wane overtime.Thisresultis con-
and significantly positivelyassociatedwiththenumberof
sistentwiththeidea thattheoutcomesof organizational
prescribedcriteriacited.These resultsprovidesupportfor
changeefforts are somewhattransient (Schein 1996,Tyre
Hypothesis2.2
and Orlikowski1994, Weick 1996), even thoughbank
managersperceivedthatstandardization effortswerecon-
Table 1 Distributionof Decision CriteriaCited by Year tinuing.Althoughseniormanagersperceiveda slightde-
cline in pressuresforstandardization over thetwo years
1986 1987 1988 1989 1990 1991 followingthestudyperiod(from5.88 to 5.63), theresults
suggestthattheirperceptions ofthedeclinelaggedbehind
% of CriteriaCited 78.0% 82.7% 91.1% 88.0% 87.6% 84.0% the materialchange in decision-making practicestaking
ThatAre Prescribed place yearsbefore.
Prescribed Criteria 2.38 3.10 4.20 4.12 3.97 3.82
Hypothesis5 predictedthatthegreatertheproportion
Ad Hoc Criteria 0.67 0.65 0.41 0.56 0.56 0.73
of prescribeddecision criteriaused, the less likely a
Total CriteriaCited 3.05 3.75 4.61 4.68 4.53 4.55
changein thedecisionassessmentfromtime(t-l) to time

Table 2 Means, Standard Deviations, and Intercorrelationsa
Variables Mean S.D. 1 2 3 4 5 6 7 8
Percentage of CriteriaThatAre Prescribed 0.57 0.75

Numberof Prescribed CriteriaCited 1.99 0.39 0.47**
NumberofAd Hoc CriteriaCited 0.15 0.20 -0.96** - 0.42**
Size of Loan 1.12 1.28 0.10** 0.11 ** - 0.06t
Durationof Relationship 14.47 14.89 0.10** 0.14** - 0.10** 0.03
SubunitSize 85.18 21.30 0.16** -0.12** -0.20** 0.00 -0.06t
CustomerBase 0.47 0.06 -0.03 -0.14** 0.01 -0.10** -0.10** 0.15**
PriorUnitPerformance 1.66 1.23 0.00 - 0.12** - 0.02 0.04 - 0.12** -0.02 - 0.19**
Strengthof Standardization 4.86 1.44 0.07* 0.25** -0.04 0.10** 0.06t 0.15** -0.15** -0.20**
aN = 900, tp < 0.10, *p < 0.05, p < 0.01
Table 3 Results of Repeated Measures Regression Analyses
Percentage of Criteria Numberof Prescribed Numberof Ad Hoc

Cited ThatAre Prescribed CriteriaCited CriteriaCited
Steps 1-2 Steps 3-5 Step 6 Steps 1-2 Steps 3-5 Step 6 Steps 1-2 Steps 3-5 Step 6
Decision- Branch- Org.- Decision- Branch- Org.- Decision- Branch- Org.-
IndependentVariables Level Level Level Level Level Level Level Level Level
Intercept 0.54 0.56 0.56 2.00 2.00 1.99 0.12 0.14 0.13
Decision-LevelVariables
Size ofthe Loan 0.09** 0.09 0.09 0.04** 0.03 0.03 -0.02** -0.02 -0.02
Durationof Relationship 0.08** 0.09 0.08 0.05** 0.05 0.04 -0.02** -0.03 -0.02
Branch-LevelVariables
SubunitSize 0.17* 0.16 0.01 -0.01 -0.05* -0.05
CustomerBase - 0.02 -0.02 -0.05 -0.03 0.00 0.01
PriorUnitPerformance - 0.01 -0.00 -0.05 - 0.03 - 0.00 -0.00
Organization-LevelVariable
Strengthof Standardization 0.02 0.09 -0.01
Total R2 0.02 0.05 0.05 0.03 0.06 0.10 0.02 0.06 0.06
IncrementalR2 0.03 0.00 0.03 0.04 0.04 0.00
ProportionofVariance Explained
At Decision-Level 0.03 0.09 0.03
At Branch-Level 0.14 0.08 0.17
At Organization-Level 0.05 0.25 0.00
N = 900, tp <= 0.10, *p <= 0.05, **p < 0.01

and heteroscedasticitycorrections.
R2 values exclude the variance explained by the autocorrelation
(t) (i.e., themorestabletheassessmentin theshortterm). consistentwithour argument, thenumberof ad hoc cri-

Consistentwiththishypothesis, theresultsshow thatthe teriacitedwas positivelyassociatedwhilethenumberof
moredecisionmakersused theprescribedcriteriain time prescribedcriteriawas negativelyassociated with the
(t), theless likelytheywouldbe to changetheborrower's likelihoodof change (p < 0.01 and p < 0.05, respec-
riskratingfromthepriorassessment(timet-1 to t) (pa- tively).All threetestssupportthecontention thattheal-
rameterestimate= -0.26, X2 = 10.24,p < 0.01). Also locationof attentionon theprescribedcriteriawouldlead
2001

to stabilityin borrowerassessments.We arguedthatsta- targetsthatcaused decision makerseitherto be lax in

bilityin ratingsover timecould benefitan organization. or to ignoreimportant
theirsearchefforts, information,
Our logic was based on the idea thatby focusinga de- and thissubsequentlyaffectedtheirjudgments.
cision maker'sattentionto the mostimportant and rele-
vant information, rule-followingcan cause a decision
makerto ignoreextraneousinformation thatadds noise Discussion
to theborrowerevaluationprocessand unnecessaryvol- The primaryfindingsof our study are that decision-
atilityto theratings.A possibledownsideto rulefollow- makingpracticeis not solely a functionof individual
ing,however,is thatof "appropriate, ratherthanoptimal, choice,and thatdecisionpracticeinfluences decisionout-
behavior" (March 1981). As March (1987) highlights, comes. Decision practicewas influencedby aspectsre-
proceduresmay be followed because they have been latedto thedecisionas well as aspectsrelatedto thecon-
learnedas appropriatein a particularsituationor as part textin which decision makerswere situated.Decision
of a particularrole, ratherthan because theyreflecta makerswere morelikelyto use prescribeddecisioncri-
deepercommitment to rationalchoice as a basis forac- teria for importantdecisions (decisions that involved
tion.Thus,thestabilityin theassessmentmayalso reflect largerresourcecommitments and unknowntargets).In
cognitiveinertiaand may foreshadownegativeconse- addition,decision makersin largersubunitswere more
quences ifthisstabilityis a consequenceof stickingwith likelyto adhereto prescribedpractice.Reliance on pre-
rules and overlookingotherimportantinformation that scribedcriteriain theshorttermwas associatedwithsta-
lies outsidetheprescribedcategories. bilityin decisions.Decision makerswho used a greater
To assess thesecompetingperspectives, we conducted proportionof prescribedcriteriawere less likely to
a post hoc analysisto examinethe relationship between changetheirriskassessmentsfromthepriorperiod.But
relianceon prescribedcriteriain time(t) and the likeli- relianceon prescribedcriteriain thelong termhad some
hood thata borrower'sriskratingwould be downgraded untowardconsequences; decision makerswho used a
in the followingyear (t + 1), reflecting increasedrisk greaterproportion of prescribedcriteriaweremorelikely
associatedwithloans made to theborrower.Researchon to downgradeloan riskin thefollowingperiods.
bank lendingpractices(McNamara and Bromiley1999) From a theoreticalstandpoint, beyondconfirming the
suggeststhatdowngradingof risk is a key indicatorof idea thatdecisionmakerbehavioris situatedand is not
decision ineffectiveness because the profitability of a simplya functionof individualchoice, this studycalls
commercialloan decreases as its risk increases.After intoquestionthe role intentional changeeffortsplay in
controllingfor macroeconomicconditions,we founda influencingindividualdecision maker behavior. Most
positiveassociationbetweenthepercentageofprescribed studiesofplannedchangefocuson summary measuresof
criteriaused in time (t) and the likelihoodthata loan job performance or changesin organizationaloutcomes
would be downgradedin thefollowingyear(t + 1) (pa- and have provided little insight into changes in
rameterestimate= 0.15, X2 = 4.50, p < 0.05) and two individual-levelbehavior. Our findingssuggest that
yearsin thefuture(t + 2) (parameterestimate= 0.31, planned organizationalchange can influencedecision
X2 = 7.05, p < 0.01). We also foundevidencethatthe makersto significantly altertheirjob-specificbehavior,
moread hoc criteriacitedin time(t), theless likelya loan at least fora time.But as timeprogresses,adherenceto
would be downgradedin timeperiodst + 1 and t + 2 newpracticesappearsto erode.The findings, in part,con-
(p < 0.10 and p < 0.05, respectfully). We foundno sig- trastwith thefindings of some studiesof planned change
nificant betweentheabsolutenumberof pre-
relationship and technologicaladaptation,but are consistentwith
scribedcriteriacited and likelihoodthatthe loan would studiesof adaptivestructuration and processinnovation.
be downgradedin time(t + 1), butwe did finda positive Plannedchangemostoftenoccursthrough thereplace-
relationship whenassessingthelikelihoodof downgrad- mentofone setofbehaviorsforanother (FordandBackoff
ing in time(t + 2) (p < 0.01). Together,thesefindings 1988). Howeversome studiesshowthatreplacement sel-
suggestthatthe extentto which decision makersover- dom works(e.g., Beer et al. 1990, Tyreand Orlikowski
looked,ignored,or otherwisefailedto seek out and con- 1994). In our study,replacementworkedinsofaras de-
sidertheidiosyncratic ad hoc criteriaand insteadfocused cisionmakerswerequickto adoptthenew decisionprac-
on theprescribedcriteriain theirevaluationsaffectedthe tice afterit was introduced,afterdecision makersre-
likelihoodthata loan riskratingwould be downgraded ceived trainingand revisedproceduremanuals,whenthe
in future timeperiods.One plausibleexplanationforthese new practicewas highlysalient.But use of prescribed
findingsis thatrelianceon theprescribeddecisionrules criteriawaned over time (while use of ad hoc criteria
engendereda myopicor automaticprocessingofdecision waxed) even thoughtop management continuedto exert

pressuresforconsistencyand conformity. This suggests makers' actions were conscious or based on insights
thatorganizationaleffortsto influencethe more micro gained throughtheirexperiencein using the prescribed
aspectsof decision makingare powerfuland yet some- criteriaovertime.
whattransitory. This may be because of the difficulties In additionto providinginsightintohow practicesare
in establishing new behaviors,because newbehaviorsof- produced,our findingslink decision outcomeswiththe
tendecay over time,or because decisionmakersregress mode of theirproduction.One of themoresurprising re-
to earlierlearnedbehaviors.Alternatively, consistent with sults of our studyrelatesto the outcomesof using the
theoriesof adaptivestructuration (see Poole and DeSanctis prescribeddecisioncriteria.Decision makerswho used a
1990, and DeSanctis and Poole 1994), technologicaland greaterproportion of prescribedcriteriawere less likely
process innovation(e.g., Barley 1988, p. 51, Tyre and to changetheriskratingfromt-1 to t. At thesame time,
Orlikowski1997),and accommodation (Orlikowski1996), decision makerswho used a greaterproportionof pre-
whichaccountforstabilityand changein organizational scribedcriteriawere morelikelyto downgradethe risk
processesand practices,thechangein theuse of criteria of theborrowerin thefollowingtimeperiods.This find-
over timecould representa processthroughwhichloan ing suggeststhatloan officerswho relied on the pre-
officers producedandreproducedtheirownversionofthe scribedcriteriamade decisions thatwere less effective
procedureforfutureuse. thanthosewho did not.One plausibleexplanationis that
Centralto the processof adaptivestructuration is the by followingthe rules,decision makerstook a shortcut
interplay betweenwhatPoole and DeSanctis (1990) have thatcaused themto restricttheirsearch,ignorediscrepant
distinguished as the spiritof the technologicalstructure information, or be automaticin theirprocessing.Perhaps
(i.e., specifictoolsandtechniquesforinstrumental action) by encouragingthejudicious use of information, theor-
and the specificfeaturesdecisionmakersuse. The spirit ganizationinadvertently encourageddecision makersto
reflectsthegeneralgoal of thetool (in ourcase, thegoal narrowtheiranalyticframeand simplyconsiderinfor-
of prescribedpracticewas to promoterationaldecision mationthatneatlyfitintotheorganization'sformalcate-
making),while the featureis a specificrule or resource gories.Perhapsby stickingwiththerules,decisionmak-
thatoperatesin a group(in our case, thefeaturewas the ers became complacentaboutdecisionsand did notfully
use-or not-of prescribedcriteria).Even thoughusers evaluatetheborrower.
or othersmayplan and tryto controltheuse of practices, Combined,the resultssuggestthatplannedorganiza-
decision makersmay appropriate(use, adapt,or repro- tionalchangecan be a powerfultool in changingtherou-
duce) practicesin ways thatare consistentwiththeorigi- tinebehaviorsof individualdecisionmakersand can in-
nal intent(faithful appropriation) or contradictory (ironic crease the level of consistencyand conformancein
appropriation) (Poole and DeSanctis 1990). And in this decisionprocessesacrossorganizational subunits, at least
way,practicesare modifiedgraduallythrougha seriesof for a period of time. The resultsalso demonstrate the
small imperceptiblechanges,which are shaped by the complexhierarchy offactorsthatinfluencedecisionmak-
context(Poole and DeSanctis 1990). Withrespectto our ers' behavior.However,theresultsprovidea cautionary
study,it is possible thattheprescribedcriteriafailedto note in thattheysuggestthatextremepressuresto stan-
capturethe rangeof information necessaryto fullyand dardizeimportant, yetrelativelycomplex,decisionsmay
effectively make a judgment.The slippagein theuse of not work as intended.The findingshighlightthe chal-
the prescribedcriteriamay reflectdecision makers'ac- lengesof designingdecisionsystems.
tionsto supplement theprescribedinformation withother One implicationfororganizations thatwantto increase
information thatwas criticalforeffectively assessingthe decision-making consistencyand predictability through
creditworthiness of a borrower.Althoughwe treatedthe formaldecisionpracticesis theneed to be mindfulof the
two setsof criteriaas substitutable, in practicetheymay untowardconsequencesthatcan occur. Our findingsdi-
be complementary. The findingsshownin Table 1 sug- rectattention to thecontextualelementsthatregulatehow
gestthatwhiledecisionmakerswerefaithful to thespirit rules are followed,established,and changed.Perhapsa
(theycontinuedto citemoredecisioncriteriain totalthan set of practicescan be designedforsome staticdecision
theycitedbeforethenew practicewas introduced), they problemsin organizations.But, as thisstudyshows,the
alteredthe features(over time, they cited fewerpre- moredifficult challengeinvolvesdesigninga systemfor
scribedcriteriaand moread hoc criteria)to accommodate more complex,imprecise,changingdecision problems.
to thetaskand context.The patternof findingssuggests Designersof decision-making systems,especiallythose
thatovertimedecisionmakersadaptedthefeaturesofthe targetedtowardcontrolling information,needto be mind-
practice(decisioncriteriaused) to makethepracticemore ful thattheycannotfullycontrolpractice,thatdecision
effective,althoughwe do not know whetherdecision makersare boundto adapttheirpracticesover time,and
ORGANIZATIONSCIENCE/VO1. 12, No. 4, July-August 2001 497

thatdecisionpracticesmustbe sufficiently flexibleto al- decision-making behaviorin otherdecisionsettings. Still,

low decisionmakersroomto adapt as necessary(March we believe thattheresultspresentedin thispaperare in-
1987). Decision makersmay make mistakeswhen they terestingand representa valuable step in investigating
follow routineproceduresin the face of changingcon- how aspectsof thenaturaldecision-making environment
ditions. Although, in theory,decision reliabilityis influencethe information thatdecision makersuse in
thought to be achievedthrough theuse offormaldecision makingimportant decisions.
rules,the singularfocus on repeatability as the primary Finally,we assumedthattheinformation decisionmak-
definingquality of reliabilityin traditionaldefinitions ers citedforeach decisionreflectedthecriteriatheyac-
fails to deal with the realitythatdecision makersface tually used in makingtheirrisk-rating judgment.Al-
fluctuating conditionsthatare not always knownin ad- thoughcrude,ourpreliminary analysissuggestedthatthe
vance. Our ideas are consistentwithothersstudyingsit- information reportedwas congruent withtheinformation
uationassessmentin complexsystems(e.g., Weick et al. thatdrovetheriskassessment.Of course,we haveno way
1999) who suggestthatproblemsolversmustbe oppor- of knowingforsurethatthecriteriadecisionmakersre-
tunisticand flexiblein theirevidencecollectionand eval- portedin theirloan recordsreflecttheactualcriteriathey
uationtacticsto detectand evaluate new eventsand to used to make theirjudgments.
revisetheirunderstanding of a situation. Several opportunitiesforfutureresearchfollowfrom
thisstudy.First,futurestudiesshouldconsiderthe role
of expertiseand experiencebecause these factorshave
Limitationsand FutureResearch been shownto be important notonlyin affecting theuse
There are fourlimitationsto this studythatneed to be of decisionpractices(expertsare less likelyto use them)
noted.First,earlierwe remarkedthatthebankspentcon- but also in influencingdecision outcomes (see Klein
siderabletime developingthe prescribeddecision pro- 1998). Second, we did not have data on the extentto
cess, and we assumed thatthe set of prescribedcriteria whichdecisionmakersreceivedfeedbackon theirdeci-
includedthe information necessaryto make a superior sion performance or theextentto whichdecisionmakers
judgment.It is possible,however,thattheprescribedcri- believedtheirperformance would be enhancedby using
teriawereinadequateforthetask.Consequently, we can- thestandardcriteria.Laboratorystudies(e.g., Arkesetal.
not definitivelyconclude thatthe adverseconsequences 1986,Davis and Kottemann1995) have shownthesefac-
associatedwiththeuse of prescribeddecisioncriteriaare torsto be veryimportant in affectingdecisionmakeruse
not simplybecause of the implementation of "poor de- of standardizeddecision practicesand the consequent
cision rules." Interviewswith top managerssuggested performance outcomes,so these factorsshould also be
thattheorganizationwentto greatlengthsto identify the includedin futurefieldstudies.Third,researchers study-
most importantand generalizablefactorsthat lenders ing groupdecisionsupportsystems(e.g., DeSanctis and
were to consider.Moreover,top management'scontinu- Poole 1994,Tyreand Olikowski1997) have studiedhow
ing commitment to the prescribeddecision process sig- practicesare changedthroughuse, but theycould build
nifiesthattheycontinuedto believe in the utilityof the on our findingsto betterunderstand how contextualele-
decision practiceseveral years afterit was introduced. mentsmightshape the structuration process.Finally,in
Thus, it is highlyprobablethatthepressuresto consider theorizing abouttheoutcomesassociatedwithprescribed
these criterialed decision makersto allocate too much practice,we discussed,butdid notempiricallyexamine,
attentionto these particularfactorsat the expense of certainfactorsthatwe argued mightmediatebetween
other,perhapsmore important but idiosyncratic,
indica- practiceand outcomes.For example,we arguedthatthe
torsof borrowerhealth. use of a prescribedpracticecould resultin cognitivein-
Second, like most studiesin the social sciences,the ertiaor automaticprocessingthatcould lead to poorer
factorsincludedin thisstudyprovideonly a partialex- outcomes;yetin our model,we did notempiricallyex-
planation of the phenomenon.Undoubtedlythere are aminethesefactors.A finaldirectionforfutureresearch,
otherimportant factors(bothindividualand contextual) therefore, would involve the articulationand empirical
thataffectthecriteriadecisionmakersuse to makerisky examinationof the specificfactorsmediatingrelation-
decisions. shipsbetweentheuse of standarddecisionpracticesand
Third,we analyzed a very specifictype of decision decisioneffectiveness.
withina singleorganization, whichmakesithardto gen- In sum,organizationsoftentryto compensateforde-
eralize our findingsbeyond this particularcommercial cisionmakers'limitedcognitiveabilitiesbycreatingstan-
lendingorganization.Futureresearchshouldexaminethe dard decision-making practices.This assumesthatdeci-
degree to which the factors we explored influence sion makerswill be rationalin usingprescribedpractices
498 ORGANIZATIONSCIENCE/Vo1. 12, No. 4, July-August 2001

and neglectsthesituationally
dependentnatureofbehav- Carter,E. E. 1971. The behavioraltheoryof the firmand top-level
ior.Our findingssuggestthatdecisionmakingis notsim- corporatedecision.Admin.Sci. Quart. 16(4) 413-429
ply a matterof individualchoice,but is shapedby mul- Cohen, J. 1960. A coefficient of agreementfornominalscales. Edu-
cational and Psych.Measurement20(1) 37-46.
tiplefactors.Our findingscan informbothscholarsand
, P. Cohen. 1983. AppliedMultipleRegression/Correlation Anal-
practitioners
who are interestedin understanding
and im-
ysisfor the Behavioral Sciences. Lawrence Erlbaum,Hillsdale,
provingdecisionmakingwithinorganizations. NJ.
Cohen,M. D., J.G. March,J.P. Olsen. 1972. A garbagecan modelof
Acknowledgments organizationalchoice. Admin.Sci. Quart. 17(1) 1-25.
The authorsareespeciallygrateful
to David Hoffman andJohnHollenbeck Davis, F., J. Kottemann.1995. Determinants of decisionruleuse in a
fortheirexpertiseand advice on data analysis.Theyalso thankBarbara productionplanningtask.Organ. Behaviorand HumanDecision
Carlin,PhilipBromiley,AmyPablo,Sim Sitkin,and members of theDe- Processes 63(2) 145-157.
cisionConsortium at theUniversity of Michiganfortheircommentson Day, D. V., R, G. Lord. 1992. Expertiseand problemcategorization-
previousversionsofthiswork.Finally,theauthors thankBill Stansifer
and The role of expertprocessingin organizationalsense-making.J.
otherswho assistedthemwithdatacollectionand providedcomments on ManagementStud.29(1) 35-47.
theirfindings.
Thisworkwas fullycollaborative. Dean, J.W., M. P. Sharfman.1993. Proceduralrationality in thestra-
tegicdecision-making process.J. ManagementStud.30(4) 587-
Endnotes 610.
'To assess therobustnessofourfindings, we conducteda supplemental 1996. Does decisionprocessmatter?A studyof strategic
analysis and testedthe hypothesizedrelationshipsusing hierarchical decision-making effectiveness. Acad. ManagementJ. 39(2) 368-
linearmodeling(Hofmann1997), butwithoutcontrolling forautocor- 396.
relation.The resultsfromthisanalysiswere fundamentally consistent DeSanctis,G., M. S. Poole. 1994.Capturing thecomplexityinadvanced
withthe resultsusing the repeatedmeasuresregressionanalysis.Re- technologyuse: Adaptivestructuration theory.Org.Sci. 5(2) 121-
sultsfromthisanalysisare available upon requestfromtheauthors. 147.
2We also testedfora possible curvilineareffectin Hypothesis2. An Dutton,J. E. 1993. Interpretations on automatic:A different view of
alternativeto therelationship
hypothesized in Hypothesis2 is thatlend- strategicissue diagnosis.J. ManagementStud.30(3) 339-357.
ers who have developed close (long duration)trustingrelationships Eisenhardt,K. M., L. J. Bourgeois III. 1988. Politicsof strategicde-
withborrowersmay be more likelyto take idiosyncraticinformation cision makingin high-velocityenvironments: Towards a mid-
about themintoaccountto justifydecisionsthatfavorthem(the bor- rangetheory.Acad. ManagementJ. 31(4) 737-770.
rowers).Thus, we testedto see if relianceon prescribedcriteriarose Feldman,M. S., J. G. March. 1981. Information in organizationsas
and thenfellas durationincreased.Our additionalanalysisoffered little signal or symbol.Admin.Sci. Quart. 26(2) 171-186.
supportforthisproposedrelationship, althoughwe are gratefulto Se- Ford, J. D., R. Backoff. 1988. Organizationalchange in and out of
niorEditorM. ScottPoole foralertingus to thispossibility. dualitiesand paradox. R. E. Quinn, K. Cameron,eds. Paradox
and Transformation. Ballinger,Cambridge,MA, 81-121.
References Fredrickson, J. W. 1984. The comprehensiveness of strategicdecision
Arkes,H. R., R. M. Dawes, C. Christensen.1986. Factorsinfluencing processes: Extension,observations,and futuredirections.Acad.
ManagementJ. 27(3) 445-466.
theuse of a decisionrulein a probabilistictask.Organ.Behavior
A. L. laquinto. 1989. Inertiaand creepingrationalityin strategic
and Human Decision Processes 37(1) 423-442.
decisionprocesses.Acad. ManagementJ. 32(3) 516-542.
Balanda, K. P., W. R. Stanton,J. B. Lowe. 1999. Predictorsof sun
Gioia, D. A., P. P. Poole. 1984. Scriptsin organizationalbehavior.
protectivebehaviorsamongschoolstudents. BehavioralMedicine
Acad. ManagementRev. 9(3) 449-459.
25(1) 28-35.
Granovetter, M. 1985. Economicactionand social structure: The prob-
Barley,S. R. 1988. Technology,power,and thesocial organizationof
lem of embeddedness.Amer.J. Socio. 91(3) 481-510.
work.N. Tomaso, ed. Researchin theSociologyofOrganization.
. 1992. Problemsofexplanationin economicsociology.N. Nohria
JAI Press,Greenwich,CT, 33-80. and R. Eccles eds. Networksand Organizations:Structure, Form
Beer, M., R. A. Eisenstat,B. Spector. 1990. Why change programs and Action.HarvardBusiness School Press,Boston,MA, 25-56.
don't producechange.Harvard Bus. Rev. 68(1) 158-167. Hedlund,J.,D. R. Ilgen,J.R. Hollenbeck.1998. Decision accuracyin
Bromiley,P. 1987. Do forecastsproducedby organizationsreflectan- computer-mediated versus face-to-facedecision-makingteams.
choringand adjustment? J. Forecasting6 201-210. Organ. Behaviorand Human Decision Processes 76(1) 30-47.
Brush,T. H., K. W. Artz. 1999. Toward a contingent resource-based Hickson,D. J.,R. J. Butler,D. Cray, G. R. Mallory,D. C. Wilson.
theory:The impactof information asymmetry on the value of 1986. Top Decisions: StrategicDecision-Makingin Organiza-
capabilities in veterinarymedicine. StrategicManagementJ. tions.Jossey-Bass,San Francisco,CA.
20(3) 223-250. Hofmann,D. A. 1997. An overviewof the logic and rationaleof hi-
Bryk.A. S., S. W. Raudenbush.1992. Hierarchical Linear Models. erarchicallinearmodels.J. Management23(6) 723-744.
Sage, NewburyPark,CA. Hollenbeck,J. R., D. R. Ilgen, D. J. Sego. 1993. Repeated measures
Carter,R. L., M. Feuerstein,J.T. Love, 1986. Repeatedmeasuresre- regressionand mediationaltests:Enhancingthe power of lead-
gression-An applicationto the analysis of mood and chronic ershipresearch.LeadershipQuart.5(1) 3-23.
pain. Comput.Statist.and Data Anal. 4(2) 131-140. J.A. LePine, J.A. Colquitt,J.Hedlund. 1998. Extending
2001

the multileveltheoryof team decision making:Effectsof feed- Northcraft, G. B., M. A. Neale. 1987.Experts,amateurs,andrealestate:
back and experiencein hierarchicalteams.Acad. ManagementJ. An anchoring-and-adjustment perspectiveon propertypricingde-
41(3) 269-282. cisions. Organ. Behaviorand Human Decision Processes 39(1)
J. Hedlund, D. A. Major, J. Phillips. 1995. The 84-97.
multileveltheoryofteamdecision-making: Decision performance Ocasio, W. 1997. Towards an attention-based view of the firm.Stra-
in teamsincorporating distributedexpertise.J.Appl.Psych.80(2) tegicManagementJ. 18 (SummerSpecial Issue) 187-206.
292-316. O'Reilly, C. A., J. A. Chatman.1996. Cultureas social control:Cor-
Huber,G. P., R. R. McDaniel. 1986. The decision-makingparadigm porations,cults, and commitment. L. L. Cummingsand B. A.
of organizationaldesign.ManagementSci. 32(5) 572-589. Staw, eds. Research in OrganizationalBehavior,Vol. 18. JAI
Jacobs,D. R., P. J.Hannan,D. Wallace. 1999. Interpreting age, period Press,Greenwich,CT, 157-200.
and cohorteffectsin plasma lipids and seruminsulinusing re- Orlikowski,W. J. 1996. Improvisingorganizationaltransformation
peatedmeasuresregressionanalysis:The CARDIA study.Statist. over time: A situatedchange perspective.Inforn.SystemsRes.
Medicine 18(6) 655-679. 7(1) 63-92.
James,L. 1982. Aggregationbias in estimatesofperceptualagreement. Phillips,J.M. 2000. Antecedentsof leaderutilizationof staffinputin
J.Appl. Psychology.67(2) 219-230. decision-makingteams. Organ. Behavior and Human Decision
Kahneman,D., P. Slovik, A. Tversky.1982. JudgmentUnderUncer- Processes 77(3) 215-242.
tainty:Heuristicsand Biases. CambridgeUniversityPress,New Poole, M. S., G. DeSanctis. 1990. Understandingthe use of group
York. decisionsupportsystems:The theoryof adaptivestructuration. J.
Klein, G. 1998. Sources of Power: How People Make Decisions. MIT Fulk and C. Steinfield,eds. Organizationsand Communication
Press,Cambridge,MA. Technology.Sage, BeverlyHills, CA, 173-193.
Landis, J.R., G. G. Koch. 1977. The measurement of observeragree- Powell, J. L. 1991. An attemptat increasingdecision rule use in a
mentforcategoricaldata. Biometrics33 (March) 159-174. judgmenttask.Organ. Behaviorand HumanDecision Processes
Langley,A. 1989. In searchofrationality: The purposesbehindtheuse 48(1) 89-99.
of formalanalysisin organizations. Admin.Sci. Quart.34(4) 598- Regehr,G., G. R. Norman.1996. Issues in cognitivepsychology:Im-
631. plicationsforprofessionaleducation.Academic Medicine 71(9)
Leblibici, H., G. R. Salancik. 1981. Effectsof environmental uncer- 988-1001.
taintyon information and decisionprocessesinbanks.Admin.Sci. Robertson,P. J.,D. R. Roberts,J.I. Porras.1993. Dynamicsofplanned
Quart. 26(4) 578-596. organizationalchange: Assessing empiricalsupportfor a theo-
Levinthal,D. A., M. Fichman.1988. Dynamicsof interorganizational reticalmodel.Acad. ManagementJ. 36(3) 619-634.
attachments:Auditor-clientrelationships.Admin. Sci. Quart. Ryan,A. M., L. McFarland,H. Baron,R. Page. 1999. An international
33(3) 345-369. look at selectionpractices:Nationand cultureas explanationsfor
Levitt,B. J. G. March. 1988. Organizationallearning.Annual Rev. variabilityin practice.PersonnelPsVch.52(2) 359-391.
Sociology 14 319-340. Seabright,M. A., D. A. Levinthal,M. A. Fichman. 1992. Role of in-
March,J. G. 1981. Decision makingperspective.A. H. Van de Ven dividualattachments re-
in thedissolutionof interorganizational
and W. Joyce,eds. Perspectiveson OrganizationalDesign and lationships.Acad. ManagementJ. 35(1) 122-147.
Behavior.JohnWiley,New York.
Schein,E. H. 1996. KurtLewin's changetheoryin thefieldand in the
. 1987. Ambiguityand accounting:The elusive linkbetweenin-
classroom:Notes towarda model of managedlearning.Systems
formationand decision-making.Accounting,Organ., and Soc.
Practice 9(1) 27-47.
12(1) 153-168.
Scott,W. R. 1987. Organizations.Rationial,Natural and Open Sys-
J. P. Olsen. 1986. Garbage can models of decision makingin
tems,2nd ed. Prentice-Hall, Inc., EnglewoodCliffs,NJ.
organizations.J. G. March and R. Weissinger-Baylon, eds. Am-
. 1995.Institutionsand Organizations.Sage, ThousandOaks,CA.
biguityand Command.Pitman,Marshfield,MA, 11-35.
Sitkin,S. B. 1992. Learningthroughfailure:The strategyof small
H. Simon. 1958. Organizations.JohnWiley,New York.
losses. B. M. Staw and L. L. Cummings,eds. Research in Or-
McNamara, G., P. Bromiley.1997. Decision makingin an organiza-
tionalsetting:Cognitiveand organizationalinfluenceson riskas- ganizationalBehavior,Vol. 14 JAI Press,Greenwich,CT, 231-
sessmentin commerciallending.Acad. ManagementJ. 40(5) 266.
1063-1088. Szulanski,G. 1996. Exploringinternalstickiness:Impedimentsto the
1999. Risk and returnin organizationaldecisionmaking. transfer of best practiceswithinthe firm.StrategicManagement
Acad. ManagementJ. 42(3) 329-339. J. 17 27-43.
Miller, D. 1993. The architecture of simplicity.Acad. Management Tetlock,P. E. 1992. The impactof accountabilityon judgmentand
Rev. 18(1) 116-138. choice: Toward a social contingencymodel.Adv. Experimenital
Mowday, R. T., R. I. Sutton.1993. Organizationalbehavior:Linking Psych.25 331-376.
individualsand groups to organizationalcontexts.Annual Rev. Thompson,J. D. 1967. Organizationsin Action.McGraw-Hill,New
Psych.44 195-229. York.
Nayyar,P. 1990. Information asymmetries:A source of competitive A. Tuden. 1959. Strategies,structures and processesof organi-
advantagefordiversifiedservicefirms.StrategicManagementJ. zationaldecision.J.D.Thompson,ed. ComparativeStudiesinAd-
11(7) 513-519. ministrationi.The Universityof Pittsburgh Press,Pittsburgh,PA.

Tyre,M. J.,W. J. Orlikowski.1994. Windowsof opportunity: Tem- Weick, K. E. 1996. Drop yourtools: An allegoryfororganizational
poral patternsof technologicaladaptationin organizations.Or- studies.Admin.Sci. Quart.41(2) 301-313.
gan. Sci. 5(1) 98-118. , R. E. Quinn. 1999. Organizationalchangeand development.An-
fortechnologicalimprove-
1997. Exploitingopportunities nual Rev. Psych.50 361-386.
ment in organizations.M. L. Tushman and P. Anderson,eds. , K. M. Sutcliffe,D. Obstfeld.1999. Organizingforhighreliabil-
Managing StrategicInnovationand Change. OxfordUniversity ity:Processes of collectivemindfulness. R. Suttonand B. Staw,
Press,New York,53-67. eds. Research in OrganizationalBehavior,Vol. 21. JAI Press,
Van de Ven, A. H., D. L. Ferry.1980. Measuringand AssessingOr- Greenwich,CT, 81-124.
ganizations.JohnWileyand Sons, New York. Worchel,S., J. W. Brehm.1971. Directand impliedsocial restoration
Walker,A. G., J.W. Smither.1999. A five-yearstudyof upwardfeed- of freedom.J. Personalityand Soc. Psych. 18(2) 294-304.
back: What managersdo with theirresultsmatters.Personnel Zucker,L. G. 1987. Institutional AnnualRev.
theoriesof organization.
Psych.52(2) 393-423. Sociology 13 443-464.
AcceptedbyM. ScottPoole.
ORGANIZATIONSCIENCENVol. 2001
12, No. 4, July-August 501


Controling

Uploaded by

Document Information

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

Controling

Uploaded by

Copyright:

Available Formats

Controlling Decision-Making Practice in Organizations

Author(s): Kathleen M. Sutcliffe and Gerry McNamara

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

Abstract and theirassociatedoutcomes.This studyinvestigates the

ORGANIZATION SCIENCE, ? 2001 INFORMS 1047-7039/O1/1204/0484/$05.OO

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

measuresofjob performance orchangesinorganizational makingpracticebyexamining theextentto whichdecision-

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

Langley 1989). An important unansweredquestionin the In summary,our studybridgeslevels of analysisby

486 ORGANIZATIONSCIENCE/VO1. 12, No. 4, July-August 2001

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

resultto constrainedvariancein theirmeasureofdecision makerlacks knowledgeaboutan exchangepartner, he or

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

the idea thatdecisionmakerstypicallydevelop context- means,theprinciplesof action,interaction, and interpre-

488 ORGANIZATIONSCIENCE/Vol. 12, No. 4, July-August 2001

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

criticalfactorsthatindicateimportant changesin thede-in the studyhad been loosely affiliatedwiththe parent

ORGANIZATIONSCIENCE/Vol. 12, No. 4, July-August 2001 489

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

criteriawere printedon a table titledthe "Risk-Rating ways. Commerciallendingdecisions involve allocation

490 ORGANIZATIONSCIENCE/VOl. 12, No. 4, July-August 2001

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

ORGANIZATIONSCIENCE/Vol. 12, No. 4, July-August 2001 491

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

of thebankbranchwas measuredusingtheaveragesize "simultaneously investigate relationshipswithina partic-

492 ORGANIZATIONSCIENCE/Vol. 12, No. 4, July-August 2001

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

responsesareaggregatedto arriveat a single,independent 5 is dichotomous.The independent variablein whichwe

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

494 ORGANIZATIONSCIENCE/Vol. 12, No. 4, July-August 2001

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

Table 2 Means, Standard Deviations, and Intercorrelationsa

Variables Mean S.D. 1 2 3 4 5 6 7 8

Percentage of CriteriaThatAre Prescribed 0.57 0.75

aN = 900, tp < 0.10, *p < 0.05, p < 0.01

Table 3 Results of Repeated Measures Regression Analyses

Percentage of Criteria Numberof Prescribed Numberof Ad Hoc

N = 900, tp <= 0.10, *p <= 0.05, **p < 0.01

(t) (i.e., themorestabletheassessmentin theshortterm). consistentwithour argument, thenumberof ad hoc cri-

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

to stabilityin borrowerassessments.We arguedthatsta- targetsthatcaused decision makerseitherto be lax in

496 ORGANIZATIONSCIENCE/Vol. 12, No. 4, July-August 2001

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

ORGANIZATIONSCIENCE/VO1. 12, No. 4, July-August 2001 497

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

thatdecisionpracticesmustbe sufficiently flexibleto al- decision-making behaviorin otherdecisionsettings. Still,

498 ORGANIZATIONSCIENCE/Vo1. 12, No. 4, July-August 2001

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

500 ORGANIZATIONSCIENCE/Vol. 12, No. 4, July-August 2001

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

This content downloaded from 128.197.27.9 on Mon, 23 Sep 2013 20:19:51 PM

You might also like