The contingency theory of management accounting and control 2980-2014

However, it was rapidly followedby othertechniquesoften lumpedtogether underthe generalheadingof strategic management accounting.. In his overview of the contingency theory of management

j o ur na l h o me p a g e :w w w e l s e v i e r c o m / l o c a t e / m a r

1980–2014

Lancaster University Management School, Lancaster LA1 4YX, UK

a r t i c l e i n f o

Article history:

Available online 12 February 2016

Keywords:

Management control package

Contingency theory

Management control system

Middle range theory

Knowledge in management accounting

a b s t r a c t

1 Introduction

Thispaperprovidesanoverviewoftheresearchonmanagement

accountingandcontrolwhichhasusedacontingentperspective.It

startsfrommy1980reviewofthetopic(Otley,1980)andseeksto

bringthisuptothepresentday.1However,thereareanumberof

featuresthatrequireclarificationtodefinethescopeofthisreview

First,thetopichasbroadenedinitsscopeoverthelastthreedecades

anditseemssensibletoincludeaspectsofmanagementcontrol

systems(MCSs)whichareusedinconjunctionwithmanagement

accountinginformationratherthanfocussingsolelyon

manage-mentaccountingtechniques.Second,managementaccountinghas

itselfchangedwithavarietyof‘new’techniquesbeingdeveloped

andpopularized,particularlyinthe1980sand1990s.Third,

orga-nizationshavechangedwithtraditionalhierarchicalformsbeing

modifiedintoflatterforms,andstrategieswhichemphasize

con-centratingonacorebusinessratherthanattemptingtoencompass

thewholesupplychainwithinasinglelegalentity.Thuscontrol

systems are increasingly required to operate across

organiza-tionalboundaries.Finally,theideaofcontingencyrequiresfurther

E-mail address: d.otley@lancaster.ac.uk

1 From this point on I will refer to my own papers in the third person and put any

personal comments in a footnote.

clarification,asitcanbearguedthatallresearchonthesetopics hastotakea‘contingency’approachasitbecomesrecognizedthat universalsolutionstoproblemsinorganizationalcontrolgenerally

donotexist

Inthe1970smanagementaccountingformedthecentrepiece

ofmanyorganizationaldecision-makingandcontrolapproaches Budgetarycontrolwasthedominanttechniqueusedandmostof theearlycontingency-basedresearchstudiesconcentratedonthe deploymentanduseofbudgets.Indeedmanyoftheearlystudies exposedtheflawsthatbudgetaryinformationpossessedwhenused

inamannerthatdidnotacknowledgeitslimitations.Morerecently non-financialperformance measureshave increasedin popular-ity and are seen as partof an overall controlsystem, together withavarietyofothercontrolapproacheswhichhavelittletodo withtraditionalmanagementaccounting.Forthispaperitseems mostappropriatetoconcentrateontheover-archingareaof man-agementcontrol systemswhere much of theresearchtakesan organizational approach.Decision-making,by contrast,tendsto takeanindividualapproachinformedpredominantlyby psychol-ogyandthisiscoveredinaseparatereviewbyHall(2016)inthis issue.However,theboundaryisnotalwaysclearassomestudies usebothindividualandorganizationallevelvariables,andfocuson topicswhichincludebothdecision-makingandcontrol,sothereis someoverlap

http://dx.doi.org/10.1016/j.mar.2016.02.001

1044-5005/© 2016 Elsevier Ltd All rights reserved.

introductionof activity-basedcosting (ABC) inthe early1980s,

althoughthisconcentratedongeneratinginformationforimproved

decision-making rather than control However, it was rapidly

followedby othertechniquesoften lumpedtogether underthe

generalheadingof strategic management accounting Although

this canbeinterpreted as anattempt bymanagement

accoun-tantstomaintaintheirpresenceatthecentreofbothorganizational

decision-makingandcontrol,itisalsoaconvenientlabelto

encap-sulate theapproaches that weredeveloped during this decade

However,thedominanceofaccountingcontrolwaschallengedin

theearly1990sbythecodificationofwhathasbecomethemost

widelyadoptedtechniqueinmodernorganizations,theBalanced

Scorecard(BSC),whichcombinedbothfinancialandnon-financial

performance measuresinto a singleintegrated framework The

scopeofmanagementcontrolincreasinglybegantoincludeissues

ofbothstrategicandoperationalcontrolthathadbeenspecifically

excluded(forreasonsofsimplicityandconvenience)byAnthony

(1965)inhisseminaldefinitionofthefield

Thecontextinwhich managementaccountingandcontrolis

practisedhasalso undergonesubstantial change.Organizations

havebecomelesshierarchicalandmanyhaverestructured

them-selvestofocusontheir‘corebusinesses’,leavingmoreperipheral

activitiestobe outsourced Thus organizationsnow tend tobe

embeddedintosupplychainsandnewformsofcontrolneedtobe

developedinthissituation(AndersonandDekker,2014).2 These

supplychainsspanseveraldifferentlegal entitieswithno

hier-archicaloversight,althoughthereisoftenonelargeorganization

thatdominatestheotherparticipants.Thesedevelopmentswere

reviewedin Otley(1994) but have continuedtochange

subse-quently.Inparticular,thegeneralbusinessenvironmenthasshown

anincreasingrateofchangeandcompetition,bothlocallyand

glob-ally,whichhascausedagreaterdegreeofuncertaintytobecome

apparent.Finally, technologicaldevelopments continuetodrive

changeatanincreasingrate,notleastinthechangestobusiness

practicewhich havebeen madeavailableby moderncomputer

technology and the internet At the very least this has led to

increasedenvironmentaluncertaintyandabreakdowninthe(often

implicit)predictivemodelsonwhichcontrolwasbased

Theideaoftheroleofcontingencytheoryisalsobeginningto

change.Whereasinitiallyitdevelopedfromtheideathatno

uni-versalsolutiontotheproblemsofcontrolwasfeasible,ithoped

thatempiricalworkwouldestablishthekeycontingenciesfrom

whichprescriptions tosuit differentsetsof circumstancecould

bedeveloped.However,researchoverthepastfourdecadeshas

comeupwithanextendedlistofpossiblysignificantcontingencies

thatarefacedbyorganizations,manyofwhichsuggest

conflict-ingrecommendations.Evenifresearchcouldbeprogressedona

muchgreaterscalethaninthepast,itisunlikelythatanoverall

contingencymodelcouldbedevelopedtosuggestoptimalcontrol

configurationsinallpossiblecombinationsofcircumstances.And

evenifwastoprovepossible,theworldwouldhavemovedonby

thetimetheresultswereavailable.Contingencythereforehastobe

consideredinamuchmoredynamiccontextthanpreviously,which

leadstotheneedtousemoreprocess-basedmodelswhich

exam-inethemechanismsofchangeandtheimplementationofmodified

formsofmanagementandcontrol

Thispaperwillthereforenotattempttoperforma

comprehen-sivereviewofallprevious‘contingency’studies,ofwhich there

havebeenanumber,mostnotablythatbyChenhall(2007)which

updateshis2003reviewandthispaperwillnotattemptto

dupli-catethedetailedworkheperformsinthatcomprehensivechapter

2

Chenhallnoted that“thetermcontingencymeansthatsomething

istrue onlyunderspecifiedconditions.Assuchthereisno ‘contin-gencytheory’,ratheravarietyoftheoriesmaybeusedtoexplainand predicttheconditionsunderwhichparticularMCSswillbefoundor whethertheywillbeassociatedwithenhancedperformance.”(p.191)

Hegoesontosuggestthatamuchwiderrangeoftheoriesmay proveuseful,encompassingeconomics(bothagencyapproaches andbehaviouraleconomics),psychology,sociologyand informa-tionscience.Healsosuggeststhat priorworkhasconcentrated

ontraditional,functionalisttheoriesandshouldmoveontouse moreinterpretiveandcriticalviewsinfuture.Thispaperwill ana-lyzeanumber ofpracticalandconceptualissuesthatappearto makeitlikelythattraditionalapproachestocontingent theoriza-tionshaveruntheircourseandtoarguethatitwillrequiredifferent approachestoprovideinsightfuland usefulexplanationsofthis complexsubject

2 What is contingency theory?

Theideaofacontingencytheoryofmanagementaccounting begantodevelopinthe1970sinanattempttoexplainthe vari-etiesofmanagementaccountingpracticethatwereapparentatthat time.Itdrewheavilyonthecontingencytheoryoforganizational structure which had been developed over the previoustwenty yearstocodifywhichformsoforganizationalstructureweremost appropriatetospecificcircumstances.Theindependentvariables usedto explainorganizational structure were oftentransferred wholesaleintotheemergingtheoryofmanagementaccountingto explainthedesignanduseofmanagementaccountingsystems, withadditionalvariablesbeingaddedastheyearsprogressed.3As Hopwood(1974b)hadpointedoutearlier,thedesignofa (man-agementaccounting)systemandthedesignofanorganizational structurearereallyinseparableandinterdependent,althoughthis importantobservationhastendedtobeneglectedovertheyears thatfollowed

In his overview of the contingency theory of management accounting,Otley(1980)specifiesthat“acontingencytheorymust identifyspecific aspectsofanaccountingsystemwhichareassociated withcertaindefined circumstancesanddemonstrateanappropriate matching.”(p.413).Thisindicatesthreeareastowhichattention needstobepaid.First,whataretheaspectsofthemanagement accountingsystemthataretobeexplained?Inparticular,arewe concernedjustwiththeexistenceofspecifictechniquesinan orga-nization,oralsowiththeextentandmanneroftheiruse?Studies havetendedtoberatherarbitraryintheirselectionofthe tech-niquestheyfocuson,withlittleconsistencybetweenonestudy andanotherbothinselectionandmeasurementofvariables con-nectedwiththeaccountingcontrolsystem.Second,howarethe definedcircumstancestobeselected?Again,althoughthe contin-gentvariablesusedbyorganizationtheoristshavebeenextensively usedhere,oftenonlyasubsetareusedinanyonestudymaking comparabilitydifficult.Finally,thedefinitionofwhatconstitutes

anappropriatematchinghascausedsignificantdifficultyoverthe years.Atitsmostsimple,existencehasbeentakenasindicating suchamatching,althoughthisassumesthatalong-run equilib-riumhasbeenachieved.Moresophisticatedstudieshaveusedsome variantoffirmperformance toindicatewhetheranappropriate matchinghasbeenfound,despitethelikelihoodthatMCSshave onlyasmallimpactonperformance,althoughthemeasuresused

3 This led to an ambiguity in the role of organizational structure which was the dependent variable in the organizational theory, but an independent variable in the management accounting theory, if it was included Evidently a risk of multi-collinearity could exist if it was used together with the common list of explanatory

prob-lematicasitcanbearguedthatperformanceisalsoanindependent

contingentvariableinitsownrightwhichcanexplaintheextentto

whichrelianceisplacedonaccountingsystemsinanorganization

However,thetotalityoftheseextendedapproachesimpliesa

verybroaddefinitionofcontingencytheory,whichledtoChenhall’s

(2007)previouslyquotedcommentthatthereareinevitablyarange

of contingency theories Indeed, it might be suggested that all

researchinmanagementaccountingisessentiallycontingent,in

thatitseekstodiscoverwhenspecifictechniquesmightbemost

appropriateforparticularorganizationsintheirspecific

circum-stances.Clearly,giventhispossibilityaoverallreviewneedstobe

bounded,andinitiallyasimpleapproachtowhatisincludedand

excludedwastakentoindicatesomeofthebroadtrendsthathave

occurredovertheperiodsince1980,4butconfinedtothosestudies

whichmakeaspecificreferenceto‘contingency’,orwhichcanbe

clearlyseenashavingsuchafocus

3 Survey 1980–2014

Aninitialliteraturesearchwasconducted usingthree steps

Firstly,apreliminarylistwasgeneratedbysearchingkeywordsin

theonlineversionsofsevenmajoraccountingjournals.Secondly,

afurthersearchwasconductedusingmajorelectronicdatabases

which provide access to mostof the other important journals

Finally,thelistwasfurthersupplementedbyexaminingthe

refer-encelistsofthekeyarticlesalreadyidentified.Thiswasintendedto

generatealistofrelevantarticlesthatwasascompleteaspossible

intheareaofcontingentapproachestomanagementaccounting

andcontrol

Theliteratureresearchbeganbysearchingarticlesfrommajor

relevant journals, namelythe following nine major accounting

journals: Accounting, Organizations and Society, British

Account-ingReview,EuropeanAccountingReview,ManagementAccounting

Research,JournalofManagementAccounting Research,Accounting

AuditingandAccountabilityJournal,TheAccountingReview,Journalof

AccountingResearch,andAccountingandBusinessResearch.Asallof

thesejournalsprovideonlineresources,articleswerecollectedby

keywordsearchesofthesupportingonlineplatform.Thekeywords

usedwere“contingency”;“contingent”;“management”;

“manage-mentaccounting”“relation”;“relationship”;“impact”;“effect”;and

“influence”.Thelastfivewordsaresearchedinordertoinclude

thoseessentiallycontingencystudieswhichdonotdirectlyclaim

tobesuch.Thetimespecifiedinthesearchfilterspannedfrom

1980tomid-2014.Thearticlesgeneratedwerereviewedbytitle

andabstracttoensurethattheywereappropriate.Assome

cross-disciplinarystudiesmaybepublishedinthejournalsofotherfields

amoregeneralsearchwithsimilarfilterconditionswasconducted

onthesevenmajorplatformswhich provideaccesstoa

multi-disciplinarycollectionofonlineresources;namelyScienceDirect

Freedom Collection; Business Source Premier; ProQuest;Wiley

OnlineLibrary;SwetsWise;JSTORandTaylor&FrancisSocial

Sci-enceandHumanitiesLibrary.Inthefinalstep;thereferencelists

inthereviewarticlesidentifiedandthemostcitedarticlesthus

listedwereusedtoaddtotheoveralllist;althoughthis

gener-atedonlyasmallnumberofadditions,itgavesomeconfidencein

thecompletenessofthelist.Thisprocessgeneratedanextensive

listwhichwasonlylightlyeditedtoexcludeitemswhichseemed

clearlyinappropriate.The finallistmaytherefore include some

itemsofperipheralrelevance;butitwasconsideredmoreuseful

toerronthesideofinclusionratherthanexclusion

4 But see Specklé and Kruis (2014) for an up to date review of recent developments

Table 1

Number of articles by journal.

Accounting, Organizations and Society 66 Management Accounting Research 57 British Accounting Review 22 Journal of Management Accounting Research 13 Accounting, Auditing and Accountability Journal 9 European Accounting Review 9 Accounting and Business Research 7

Information and Management 4 Strategic Management Journal 4 Administrative Science Quarterly 3

Insummary,thenumberofarticlesfromeachjournalisshown

inTable1,withonlytheaccountingjournalswhichcontainmore than four articles being included.5 It is notable that, although

a reasonable number of authors were located in the USA, the mainjournalsinwhichthisworkwaspublishedareEuropean-or Australasian-based,ratherthanUS-based.Thisprobablyindicates theUSjournalpreferenceforquantitativeworkusingsubstantial (oftenpubliclyavailable)datasetsand theuseofsophisticated quantitativemodels.Mostworkinthecontingencyfieldrequires datatobecollectedindividually(ratherthantherebeingany pub-liclyavailabledatasets)whichrestrictssamplestoquiteasmall size(rarelymorethanthelowhundreds),andthusprecludesthe extensiveuseofthemoredatahungrystatisticaltechniques.6

The total number of articles included in the bibliography amountedto236.Ofthese157couldbecategorizedasbeing empir-icalandusingquantitativemethods7;theremainderwereeither theoretical(39),review(20), qualitative(10),ormethodological (10),althoughseveralarticlescouldhavebeenincludedinmore thanonecategory.Inthe1980s,lessthan5relevantarticleswere publishedeachyearalthoughthisfigureincreasedtoaround10in thelate1990s.Despiteabriefdropofpublicationsinyears2000 and2001,thestatisticsshowagenerallyupwardtrend,asshown

inFig.1below,andindicatethesteadilygrowingamountofwork

onthistopicovertheperiod.However,thisincreasemaybepartly duetotheincreasednumberofjournalsavailable,mostnotablythe launchofManagementAccountingResearchin1990

4 Initial overview

Thereareanumberoffeaturesthatcharacterizethebulkofthe workreviewed.First,itisgenerallynotsystematic,andthe vari-ables(ormeasures)selectedforanalysisinanyparticularstudy generallydonotexactlycorrespondtothoseusedinpriorwork Thus,cumulativeresultsarerare.Second,viewingthisfeaturefrom

amorepositivestance,ithasclearlyseemedbettertoresearchers

toextendtheirworktoawiderrangeofareasinordertomapout theboundariesofthefieldratherthantoconcentratefurtheron thosetopicswhichhappentohavebeenstudiedinthepast

How-5 Just four relevant articles were found in each of The Accounting Review and the Journal of Accounting Research in the period covered.

6 Recent developments in statistical techniques now permit smaller sample sizes

to generate reliable results, and may allow more useful statistical work to be con-ducted in this field in future.

7 The majority of the quantitative empirical studies involved data collection by

Fig 1.Number of articles on the contingency theory of management accounting by year.

ever,theseextensionshavetendedtoconcentrateonincreasingthe

numberofindependentvariablesstudiedratherthanonthe

depen-dentvariablesofMCSdesignanduse,withonlyafewexceptions

Thisisconnectedwiththealmostcomplete absenceof

replica-tionstudies,perhapsaccentuatedbecausejournalsmayhavebeen

unwillingtopublishthem.Ifsuchstudiesweretofindnullresults

itislesslikelythattheywouldbeseeninprint,althoughjournal

editorstendtodisputethis interpretationdespitetheempirical

evidence.Third,limitedattentionhasbeenpaidtothe

character-izationofthemanagementcontrolsystemitself.Theinitialfocus

wasoftenonbudgetarycontrol,reflectingitswidespreadusein

the1980s,buthasbeenextendedtoexamineawiderrangeof

per-formancemeasuresandcontroltechniques,butonlyonasporadic

basis.Finally,thepredominantresearchmethodhasbeentheuseof

surveyquestionnaires(oftenatarm’slength)withquitelimiteduse

madeofcombiningthisworkwithmoreinterpretativeorcritical

qualitativemethods.Thisseemstohavebeenexcessivelylimiting

atwhatisstillanearlyandexploratorystageofthisfield

4.1 Independentvariables

More and more variables continue to be examined Major

independent variables can be grouped into external variables

andinternalvariables.8 Themostcommonlyexaminedexternal

variablesincludetechnology,marketcompetitionorhostility,

envi-ronmentaluncertainty and national culture.Themajor internal

variables areorganizational size,structure, strategy,

compensa-tionsystems,information systems, psychologicalvariables(e.g.,

toleranceforambiguity),employees’participationinthecontrol

systems,market position, productlife-cycle stage, and systems

change

Intheearlystageofcontingencyresearch,manystudies

exam-inedtherelationshipofjust oneindependentvariablewithone

dependentvariable.However,Fisher(1995)arguedthatitis

essen-tialtounderstandtheinteractionsbetweenmultiplecontingent

and control factors in determining theeffectiveness of control

systemdesign.Inotherwords,theunderstandingofthe

interrela-tionshipbetweenmultiplecontingentindependentvariablesmay

leadtoabetterframeworkfordependentvariablesdeterminant

analysis.Healsoclassifiedcontingency studiesonmanagement

controlbythelevelofanalysiscomplexity.Theworkonthe

con-tingencytheoryofmanagementaccountingcanbeclassifiedinto

3levelsofanalysisonthisbasis.Atthefirstlevelofanalysis,one

contingentindependentvariableiscorrelatedwithonedependent

variable.Atthesecondlevelofanalysis,thejointeffectofmultiple

8 The view of the author(s) has been accepted in this regard, although there is a

debate as to whether an internal variable which can be affected by the firm itself

should be treated as a contingent variable or as a dependent variable, most notably

contingentvariablesononedependentvariableisexamined.Some

ofthesevariablesmaybemoderatingormediatingvariablesrather thanindependentvariables.Thethirdlevelofanalysisexamines theeffectofthefitofmultipleindependentvariablesonseveral dependentvariables,althoughthishasbeenrelativelyrare.There

isatrendinthisstreamofresearchformorevariablestobe exam-inedinanyonestudywithincreasingattentionbeingpaidtotheir interrelationships

4.2 Dependentvariables The most widely examineddependent variables are perfor-mance,performancemeasures,budgetingbehaviour,management controlsystemdesignanditsuse, effectiveness,jobsatisfaction, changeinpractices,andproductinnovation.Performance, effec-tivenessanddesignofsystemsarethemajordependentvariables usedwithfinancialperformancebeingthemostcommonlyused outcomevariable.Oneofthemajorreasonsisthatfinancial perfor-manceisawidelyusedmeasureinmostorganizations.Moreover, mostvariablecompensationsystemsusea measureoffinancial performanceastheindicatorforincentivepayments.However,the over-relianceofperformance measureonfinancialperformance may produce biased results The study by Coates et al (1992) indicatesthatdifferentcountrieshavedifferentpreferencesfor per-formancemeasurementsystemswhichmayresultfromvarying emphasesonfinancialstability Specifically,theemployment of non-financialperformancemeasureswasmoreprevalentinU.S companiesatthattime.Manyoftheperformancemeasurements whicharehighlyvaluedandofsignificantimpactarenon-financial measurementssuchasproductionprocessmeasures,defectrates, cycletime and customer servicemeasures; or qualitative mea-surement,suchascustomers’ perception,attitudeofemployees towardsjobs,andproductinnovation.(Otley,1999).Despiteofthe importanceofnon-financialandqualitativeperformance measure-ment,thesevariableshavebeenrelativelyneglectedinthestudies reviewedhere

Thesemainlyquantitativestudiesformthecoreofwhat has becomeknownasthecontingencyapproachandhaveformedthe foundationofourknowledgeaboutthemajorvariablesthoughtto affectthedesignanduseofMCSs.Theyhavealsoincreasinglypaid attentiontotheinteractionsbetweenboththeindependentand thedependentvariables,atopicfurtherdiscussedinSections5.5 and5.6.However,itisperhapsunfortunatethattheterm ‘contin-gency’hasnowbecomeassociatedonlywiththemethodstypically employedinthisstrandofresearchwhenawiderrangeofresearch approaches are likely to giveadditional insights Whilst recog-nizingtheknowledgethathasbeengainedfromthesestudiesit seemslikelythatfurtherprogressismostlikelytobeobtainedfrom deployingamuchwiderrangeofresearchapproaches,giventhe complexnatureofthephenomenabeingstudied

difficulttocategorize,rangingfromdetailedanalysesofvery

spe-cificand smallscaleoperations,through tooverviewsof whole

areas, oftenwith a theoretical component Theempirical

tech-niques involved were usually interviews as the main evidence

gatheringtechnique,althoughthiswassometimessupplemented

bydocumentaryevidenceandquestionnaireresponsesfrom

man-agers.Thesewillnotbereviewedseparatelyindetailastheyare

fewinnumberanddifficulttocompareandaggregate,although

theyhaveyieldedsomesignificantinsightsintoMCSdevelopment,

modificationanduse.9Theworkreviewedheredoesnotdofull

jus-ticetothenumberofcontributionsalreadymadetomanagement

controlresearchusingqualitativemethods,asmanyofthestudies

wouldnotdescribethemselvesasbeingwithinthe‘contingency’

tradition.Itmaywellbethattheuseofthislabelisnow

becom-ingobsoleteasstudiesfocusonhowMCSsoccurintheformsthat

areobservedandtheeffectsthatdifferentconfigurationshaveon

managerialandorganizationalbehaviour

Ratherthantryingtoimposewhatwouldlikelybeanartificial

orderoftherangeofworkbeingconsidered,thisreviewwilltake

atopic-basedapproachandexaminetheareaswheresignificant

amountsofworkhave beenconducted,witha viewtodrawing

outsomelessonswhichcanbelearnedfortheconductoffuture

research.Thisisinevitablyselective,buthasbeendoneinthehope

thatanoverviewofafew areaswillprovemoreinsightfulthan

amorecomprehensiveyetinevitablymoresuperficialreviewofa

largenumberofwidelyvariedstudies.Thissectionwillalsoinclude

adiscussionofsomemethodologicalissueswhichhavebeen

dis-cussed.Itwillconcludebysuggestingissuesthatfutureworkmight

usefullyconsider.Afullbibliographyofallthearticlesconsidered

isattachedinAppendixA,withthereferencelistattheendofthis

articlecontainingonlythosereferencesnotincludedthere

5 Review of some selected bodies of work

Ashasbeenobservedabove,theempiricalworkwhichhasbeen

undertakenisdifficulttocategorizeneatly,withlittlecumulative

progresshavingbeenmade,duetosuccessiveauthorsdesigning

theirstudieswithonlylimitedconnectionstopreviouswork.Itis

alsoworthnotingthatthecontextwithinwhichmanyoftheearlier

studieswereconductedhaschangedsubstantiallysincethattime,

andthattheirresultsmaynotbeapplicabletotoday’s

organiza-tions.However,anumberofmajorthemeshaveemergedwhich

willnowbediscussedinmoredetail

5.1 Relianceonaccountingperformancemeasures(RAPM)

Earlyworkoncontingencytheorytendedtobebasedonprior

findingsfromorganizationalbehaviour,butoneimportantstream

developedfromconflictingempiricalobservationsthatrequired

explanation.ThisstartedwiththeworkofHopwood(1972,1974a)

whoexaminedtheeffectofseniormanagersgivingahighpriorityto

accountingperformancemeasuresinevaluatingtheperformance

oftheirsubordinates,astylelatercharacterizedasRAPM.Themain

resultwasthatwhereaccountingperformancemeasures(in

partic-ular,meetingthebudgetfortheunit)weregivenhighersignificance

thanothercriteria,suchaslong-runeffectiveness,thenavarietyof

undesirableconsequencesfollowed.AsubsequentstudybyOtley

(1978)deliberatelychoseasituationwherehebelievedthisresult

wouldnothold(i.e.,whereoperatingunitswerelargely

indepen-dentofeachother,incontrasttoHopwood’shighlyinterdependent

units),andfoundtheconflictingresultshepredicted.Heargued

9

thatthiswasthereforeacontingentresultdependentuponthe con-tingencyofsub-unitinterdependence.However,thisprovedtobe

acontentiousconclusionastherewereseveralotherdifferences betweenthetwostudies(e.g.,USversusUKorganization;private versuspublicsector;costcentresversusprofitcentres;sizeof orga-nizationalunitetc.)andsubsequentworksoughttoexplicatesome

ofthesepossibledifferencesandwhattheirconsequencesmight be

Theworkwhichfollowed,despitebeingdescribedas“theonly organizedcriticalmassofworkinmanagementaccountingatpresent” (BrownellandDunk(1991)(p.703))hadamajordeficiencyinthat verydifferentmeasuresofthecentralconcept(RAPM)wereused

indifferentstudies.Thesedifferencewereoutlinedandreviewed

in OtleyandFakiolis (2000)which concludedthat manyof the studieswerenotcomparablewitheachotherduetothese differ-encesinmeasurement.Significantly,bothHopwood’sandOtley’s studies made significant use of interviews in developing their questionnaire-based measures and in interpreting theirresults, whereaslaterstudiestendedtousearm’slengthgeneralpurpose questionnaires,sometimesusedacrossavarietyofdifferent organi-zations.Usingtheover-archingtermofRAPMwasitselfmisleading

as itcovered the useofseveraldifferent measureswhich each focussedondifferentaspectsoftheunderlyingphenomena Fur-ther,theappropriatenessofthequestionnaireitemstothespecific organizationsbeingstudiedisunclearasvalidationthrough field-workwasusuallyabsent

Inaddition,theworldhad movedonfromthesituationthat existed in the early 1970s (see Otley (1994)) for an analysis

at that date).Budgetary control had reduced in significance as

aperformance measure,withnon-financialmeasuresincreasing

inimportance.Organizationalstructureshadbecomeflatterand inter-organizationalrelationshipshadincreasedinimportanceas out-sourcinghadbecomemoreprevalent.Thesetrendshave con-tinuedtothepresentdayanditseemsunlikelythatresultswhich werefoundinthepastcanbereliedupontoholdinthepresent day.Attheveryleast,this needstobeinvestigatedratherthan reliancebeingplacedonworkconductedlongago.Theonemajor attempttoconstructabroaderframeworkwasSimons(1995)work

onLeversofControlwhichwasbasedonalargenumberofcase studiesconductedinUScompanies,generallyinvolvingdatafrom CEOsandotherseniormanagers.Hefocussedonamuchwiderset

ofcontrolsthanmostpreviousstudies(categorizedintodiagnostic, interactive,beliefsandboundarycontrols)althoughhis formula-tionhasprovedtobesomewhatproblematicasidentifiedandthen improvedbyTessierandOtley(2012).Butsubsequentresearch focussedalmostentirelyonhisdistinctionbetweendiagnosticand interactiveusesofcontrols,andtheambiguityinthedefinitionof interactiveusehasalsoledtothetermbeingusedtocovera vari-etyofdifferentmeasures.Indeed,Bisbeetal.’s(2007)reviewofthis conceptidentifiedfivedifferentdimensionsoftheconceptwhich hadbeenaggregatedinavarietyofways.Theconsequenceisthat, whentheliteraturereferstotheuseofinteractivecontrols,very differentconceptsmaybefoundwithinthissinglelabelbecauseof thedifferentmeasurementinstrumentsbeingused

However,the underlyingideabehind theearlystudies does appeartohaveacontinuingsignificanceinaworldwhichis becom-ingincreasinglydominatedbyperformancetargetsofalltypes.That

is,managerswhoplacerelianceonmotivatingappropriate subor-dinatebehaviourbysettingquantitativeperformancetargetsand emphasizingtheseaboveallelse,shouldnotbesurprisedifsuch targetsareattainedbyavarietyof(oftenunobserved)behaviours thatareoftendysfunctionaltotheachievementofoverall organiza-tionalobjectives.Thereisoverwhelmingempiricalandanecdotal evidencetothiseffectfromboththeprivatesector(e.g.,call cen-treoperativescuttingoffcallsbeforetheyreachthetargettime fordealingwithenquires)andthepublicsector(e.g.,school

theperformancetarget;hospitalambulancesbeingmadetowait

outsideA&Edepartmentssoasnottostartthetimeclockfor

tar-gettreatmenttimesetc.).Evenuniversityacademicshavebeen

affectedbythe Research AssessmentExercise in theUK where

administratorshaveconvertedthesubjectiverulesinvolvingpeer

judgementtoquantitativetargetsofachievingagivennumberof

articlesinjournalsbelievedtohaveahighqualityrating

It is therefore a pitythat this stream of researchseems to

havedriedupbecauseofthemeasurementdifficultiesassociated

withaparticularsetofmeasuresdesignedtomeasureRAPM.The

underlyingconceptstillseemssoundandveryrelevantto

mod-ernorganizations,particularlyifmodifiedtoincludenon-financial

aswellasfinancialperformancemeasuresandtargets.In

partic-ular,RAPMfocussedonthewaysinwhichlinemanagersactually

usedtheaccounting(andother)informationsuppliedtothemin

theprocessofholdingtheirsubordinatesaccountablefortheir

per-formance.Thisisstilla keyareathatisdeservingofcontinuing

attention

5.2 Environmentaluncertaintyandhostility

Ofallthevariablesusedintheoriginalstudiesofthe

contin-gencytheoryoforganizationalstructure,theonethathasgained

byfarthewidestattentionintheareaofmanagementaccounting

isthatofenvironmentaluncertainty.Thisseemstobeforseveral

reasons.First,itproducedsomeofthestrongestresultsinearly

studies,inasimilarwaytothegroundbreakingworkofBurnsand

Stalker(1961)intheorganizationalliterature.Ifanorganizationor

unitisfacedbyhighlevelsofuncertaintyitrequiresflexibleand

adaptablesystemstomanageactivitieswhenunexpectedevents

occur.Second,itcanbearguedthatenvironmentaluncertaintyhas

increasedovertheyears,inpartduetotheemergenceoftheglobal

economyandmoreextensivecompetition.Butitisalsocausedby

organizationsceasingtoattempttocontrolallaspectsofthevalue

chainwithinoneoverall (holding)organizationandtherelative

demiseofthedivisionalizedorganization.Thisexposesthesmaller

andmorefocussedorganizationalunitstomoreuncertaintythat,in

thepast,maywellhavebeenbufferedbytheoverarching

organiza-tion.Third,thereisalsotheissueofmeasurement.Herethemost

commonlyusedvariableisameasureofperceived

environmen-taluncertaintywhichcaneasilybeincorporatedintointerviews

orquestionnairesadministeredtoindividualmanagers.Although

thismaybeseenasusingaverysubjectivemeasure,itcanalsobe

arguedthatthisisthemostrelevantaspectofuncertainty—itisthe

uncertaintyperceivedbyindividualsthatwillmostdirectlyaffect

theirbehaviour.However,itisonlyindirectlyconnectedwithmore

objectivemeasuresofriskanduncertaintyandthefactorswhich

causesomeindividualstoperceivegreateruncertaintythanothers

arenotspecified.10

Associatedwith,butdistinctfromuncertaintyareotheraspects

oftheenvironment,includinghostility, particularlythat related

tointensecompetition.Althoughhostilitymayproducea

signif-icantdegreeofuncertainty,thefindingsofmoststudiesindicate

that the two features have diametrically opposed impacts on

MCSdesign anduse Hostilityhasbeenshowntobeassociated

withagreaterrelianceonaccountingcontrols(especiallybudgets)

whereasuncertaintyisassociatedwithamoreflexiblestyleof

con-trolthatisopenandexternallyfocussed.AsChenhall(2007)has

pointedout,itisanopenquestionhowthetensionbetweenthese

twofactorsshouldbemanagedinanMCSgiventheymayoften

occursimultaneously.Thisisindicativeofacommonproblemwith

10 A useful review of the impact of uncertainty on performance measurement and

contingencystudiesin determininghowmultiplecontingencies havingconflictingrecommendationsshouldbecombined.Italso drawsattentiontothecommonissueofdistinguishingbetween theexistenceofformalcontrolsandthewaysinwhichtheyare usedbymanagers

5.3 Strategy Strategyhasbeenhypothesizedtoaffectcontrolsystemsdesign

inanumberofstraightforwardways,dependingonwhich cate-gorizationofstrategyisused.FollowingMilesand Snow(1978) whodescribedfourorganizationaltypes(defenders,prospectors, analyzers, and reactors), it has been suggested that defenders undertakelittleproductinnovation,whereasprospectorsare con-tinuallysearchingfornewmarketopportunities,eitherforexisting

ornewproducts.Theformertypethereforeconcentrateonfinance, productionandengineering,comparedwithanemphasison mar-ketingandR&Dinthelatter.Analyzersarethoughttocombinethe strongestcharacteristicsofbothdefendersandprospectors,whilst reactorsareseenasanunsuccessfultypeandarerarelyexamined

In a similar vein, Porter (1980) distinguished three generic strategies—costleadership,differentiationandfocus.The competi-tiveadvantageofcostleadersisfocussedontheeconomiesofscale, theoperationof efficientprocurement andefficient production technology.Bycontrast,adifferentiationstrategyseekstofocuson providingproductsthatarehighlyvaluedbycustomerswhoare thuspreparedtopaypremiumprices.Finally,afocusstrategy ded-icatesitselftocateringforasegmentofthemarketthatispoorly servedbycompetitors.Hereagainthecontrolsystemfocuswillbe

oneithercostcontroloronproductqualitiesbasedoncustomer satisfaction

Finally,GuptaandGovindarajan(1984)developedalifecycle approachbasedonthesequenceofstrategicmissionsfrombuild, hold,harvestanddivest.Intheearlystagesthefocusison build-ingamarketshareandstrategicposition,withlesseremphasison short-termearningsorcashflow.Asthebusinessmovestoa har-veststrategyafirmceasestoinvestexceptwherestrictlynecessary andseekstoachieveshort-termprofitandcashflowratherthan increasingmarketshare.Aholdstrategyisintermediatetothese two,whereafirmseekstoprotectmarketshareandstrategic posi-tionwhilstobtainingreasonablereturns.Finally,thediveststrategy occurswhenabusinessdecidestoexitfromaspecificareaof activ-ity.Eachstrategyagaintendstobeassociatedwithaspecificgroup

ofcontrols

Although these different characterizations of strategy share somesimilarcontrolsystemattributes,theredoesnotseemtohave beenanycoherentcomparisonbetweenthem.Rather,the empir-icalstudiesontheimpactofstrategyonMCSdesignandusetend

toselectonetypologytoworkwith,andtofollowtheimplications

ofthisforMCScharacteristics,althoughagainfocussingon differ-entMCSfeaturesindifferentstudies.However,thestreamofwork

bySimons11focusesonthechoicecompaniesmakebetweenthe diagnosticandinteractiveuseofcontrolstosupporttheir strate-gicintent.Unlikepreviousempiricalstudies,inSimons’workthe content of the strategy is not seen as critical to understanding thenatureoftheinteractionbetweencontrolsandstrategy,using histypologyofdiagnosticandinteractiveuse,beliefsystemsand boundarysystems

Agoodreviewoftheearlierworkonstrategyandcontrolcan

befoundinLangfield-Smith(1997)althoughthetopicseemsto havewanedsomewhatininterestsincethattimeandnofurther

11 These are summarized in Simons (1995) Interestingly his individual studies were not identified in the literature search, despite being reported in journal articles,

beensomemoredetailed,qualitativestudiesthathavetriedto

cap-turethedetailsofactualstrategiesadoptedandthewaysinwhich

thesehave influenced theuseofcontrol systems.One ofthese

byAdlerandChen(2011)indicatedtheexistenceofa

confronta-tionstrategy (asdistinctfromcostleadershipordifferentiation)

whichwasassociatedwithcollaborativeorganizationalcultures,

lean organizational structures and practices, and training and

developmentprogramesthatfocussedondevelopingempowered,

multi-skilledteamsofself-governingandcoordinatingemployees

Thusthecontingencyworkontherelationshipbetween

strat-egyandMCSdesignanduseisfragmentedinseveralways.First,

thereisarelianceonsimplegenericstrategycharacterizationsthat

mayfailtocapturethecomplexityofrealorganizationalstrategies

Inaddition,itisnotalwaysclearwhetherwhatisbeingmeasuredis

anespousedorarealizedstrategy,orevenwhetheranoverall

orga-nizationalstrategyisbeingassumedtoapplyinanorganizational

sub-unitthatmayberespondingtodifferentlocalcircumstances

Second,thedimensionsofthecontrolsystemsexaminedarevaried

withnoagreedstructureorcontent,leadingtopiecemealresults,

althoughSimons’workrepresentsadistinctapproachgrounded

in case-basedresearch.However,it hasnotachieved anygreat

popularityincontingencywork,exceptforhisdistinctionbetween

diagnosticandinteractiveuses,perhapspartlybecauseitdoesnot

connecteasilywithanyotherestablished framework.Although

seriousattemptshavebeenmadetoachieveagreaterdegreeof

coherenceinframeworkspecificationbyMalmiandBrown(2008)

andbyFerreiraandOtley(2009)thereismuchstilltobedoneto

developausefulgeneralframeworkthatwillallowtheresultsof

futurestudiestobevalidlycompared

One recent piece of work by Klassen (2014) has suggested

thattheremaybeanunderlyingvariablethatislogicallypriorto

strategy,namelyvaluelogic.Essentially,avaluelogicisthebasic

businessmodeladoptedbyanorganizationandhasbeen

catego-rizedintothreetypes:valuechain,valuenetworkandvalueshop

(StabellandFjeldstad,1998).Avaluechainisthetraditional

struc-tureoccupyingonepositioninachainofactivitiesstretchingfrom

rawmaterialthroughtofinalproductsorservices(e.g.,a

manufac-turingcompany);avaluenetworkreliesontheconnectionsthat

anorganizationhaswithotherorganizationstoprovidean

over-allservicetoclients(e.g.,abank);avalueshopisanorganization

thatcontainswithinitarangeofspecialized resourcesthatcan

bemobilizedtoprovideapersonalizedbespokeserviceforclients

(e.g.,aprofessionalservicesfirm).Klassenhasshownthatthe

dif-ferencesbetweenthecontrolsystemsofvaluechainsandvalue

shopsarequitedistinct,andvaluelogicmaythusbemore

fun-damentaltoMCSdesignthanstrategicpositioning.Hisresultsare

lessclearforvaluenetworks(whicharegenerallyinan

intermedi-ateposition)perhapsbecausethesetakeseveralformsandamore

precisedefinitionoftheircharacteristicsmayneedtobedeveloped

ThecontingencyworkontheimpactofstrategyonMCSisthus

fragmentedanditisdifficulttofindcumulativecontributions.This

ispartlyduetoarbitraryclassificationsofstrategy,whichalthough

necessaryfor surveywork,maycategorizeactualstrategiestoo

simplistically.Herethepursuitofworkthatfocussesonthe

under-lyingvariableofvaluelogicmayprovidearoutetofurtherprogress

atthemoregenerallevel,withmoreinterpretativestudies,which

willinevitablyfocusonsmallsamplesoforganizations,providing

morespecificdetail.Thelackofcumulativeprogressisalsodueto

alackofconsensusonthecharacterizationoftheMCSitself,which

willbedealtwithinSection6

5.4 Culture

Therehasbeenasignificantamountofworkwhichexamines

theeffectofculture(usuallynationalculture)onthedesignand

operationofcontrolsystemswithsome25articlesinthe bibliog-raphybeingonthistopic.Muchoftheworkhasbeenconducted

byUS-basedauthorsstudyingChinese(orTaiwanese)cultureand manyarticlesrelyuponHofstede’searlyworkonthetopicwhich wasreportedmostfullyin 1980.Itis relevanttonotethat the database for his work consisted of IBM employees working in manydifferentlocations,somaygiveabiasedrepresentationof employeesinotherorganizationsdespiteitsimpressivenational coverage.Hofstede(1980)initiallyidentified fourmajor dimen-sions onwhich nationalculturescouldbedistinguished:power distance,individualism,masculinityanduncertaintyavoidance.He lateraddedshort-termvs.long-termemphasis,andHofstedeetal (2010)added pragmatismand indulgence basedonWorld Val-uesSurveydata.Essentially,thesestudiesdependeduponforms

ofstatisticalanalysissuchasfactoranalysistoidentifyunderlying clustersofvariableswhichwerethengiventheabovelabels What is notable about the MCS work is that differences in nationalculturesareoftenmerelyassumedtobedifferentin stud-ies,ratherthanbeingmeasuredbythem,andtheyarealsoassumed

toaccordwiththeresultsofHofstede’searlystudies,despitethe ageofhiswork.InafewcasesaninstrumentbasedonHofstede’s workisusedtoconstructameasureofsomeculturalvaluesfor

aspecificsample.Theoutcomeisthatstudiesarerarely compa-rableevenonhowcultureisdefinedorassumedtobe,letalone withitseffectsoncontrolsystems.Heretheextantworkstudies differentaspectsofcontrolsystems,withmanyfindingthat appar-entlyintuitivehypothesesarenotsubstantiated.Soitprovesalmost impossibletogeneralizeabouteventhemajoreffectsof(national) cultureonMCSdesignanduse

Thisisprobablymainlyduetotwoeffects.First,national cul-ture,evenifcapturedbythemeasurementinstrumentsused,is

anaveragewithinwhichthereisalmostcertainlyawiderangeof variationbetweenindividualsand groupswithinacountry.The employeesofaparticularcompanymaywelldiffersignificantly fromthestereotypicalbehaviourthatanoverallaverageprovides Second,itislikelythatorganizationalculturewillalsohavea signif-icantinfluenceonattitudesandbehaviourwithinanorganization Unlesssomeattemptismadetomeasuretheculturalvaluesand assumptionsoftheparticularemployeesinquestion,thenitislikely thatanyresultswillcontainsignificantamountsofnoise.Again, thereislikelytobeconsiderablevariationbetweenindividualsand,

inaddition,inthesedaysofglobalemployment,manyemployees

ina singlelocationmaywellhavebeenbroughtupindifferent nationalcultures.Butorganizationalculturecanalsobemanaged

toanimportantextentandsignificantexamplesexistoftraining regimesthathave changedthebehavioursof keyemployeesin certainorganizations(e.g.,airlinepilotsinaKoreanairlinewho previouslytendedtohavesuchahighlevelofhierarchicalrespect thattheyfailedtoquestionthedecisionsofa seniorpiloteven whenthesewerethoughttobeclearlyinappropriate,andinsome casesledtomajoraccidents).Organizationalcultureisprobably bestseenasamediatingvariablethroughwhichnationalculture actsininfluencingbehaviourinorganizations

Thesecommentsshouldnotdiminishtheimportanceof study-ingorganizationalcultureanditsimpactonemployeebehaviouras organizationalcultureisanimportantcontrolmechanism(Ouchi,

1979).AccordingtoFisher(1995)organizationalcultureimpliesa setofsocialnorms,valuesandbeliefsthataresharedbythe mem-bersoftheorganizationandinfluencetheiractions.Hearguesthat

astronginternalculturecandecreasetheneedforothercontrol mechanisms,andmaythusaffecttheoveralldesign ofanMCS Thisisthereforeanimportantareaforcontinuedwork,despitethe difficultiesthatwillneedtobeovercometoobtainresultsofany generality

5.5 Effectiveness

Theideaofacontingencyfitrequiressomemeasureof

effective-nesstoactasacriterionvariable.Thatis,agoodcontingentfitcanbe

definedasachievinghighereffectivenessthanapoorfit.Buthow

effectivenessisdefinedandmeasuredvariesconsiderablyacross

studies.Initiallysomestudiesmerelyassumedthatexistencewas

itselfameasureofeffectiveness,generallyonthe(oftenunstated)

groundsthatinthelong-runequilibriumconsiderationswouldlead

toobservedoccurrencesstemmingfromagoodfit,whereascasesof

apoorfitwouldhavegoneoutofexistence.Notonlyisthis

implau-sibleinmodernconditions,butempiricalstudieswhichanalyzea

largedatasetrequiresomevariabletoactasacriterion,soeach

suchstudyrequiresameasureofeffectiveness.Thesevary

depend-ingonthetopicofthestudy;forexampleproductinnovationhas

beenusedinstudieswherethefocusisonMCSdesignanduse

thatsupportinnovation(seeBisbeandOtley(2004)).Butthemost

commonlyusedmeasureofeffectiveness,byfar,hasbeenfinancial

performance,usuallyprofitorreturnoninvestment

Therearetwomajorproblemswithusingprofitabilityasa

cri-terionvariabletodeterminecontingencyfit.Firstly,profitabilityis

affectedbyahugerangeoffactorsotherthanMCSdesignoruse

Thusagreatdealofrandomnoisecanbeexpected.Additionally,

theunitwhoseprofitabilityisusedshouldideallybedifferentfor

eachofthemanagersincludedinthestudy(orissuesaboutthe

appropriateunitofanalysiswillberaised).Secondly,financial

per-formancehasbeenarguedtobeacontingentvariableinitsown

right.For example,anorganizationfacingpoorfinancial

perfor-manceishighlylikelytopaymoreattentiontofinancialcontrol

techniquessuchasbudgeting.Ifthisisthecase,thentheapparently

perverseresultthattheextensiveuseofbudgetingispositively

associatedwithpoorerfinancialperformancecanbeexpected

Cer-tainly,itisnotpossibletousefinancialperformanceasacriterion

variableifitiswishedtostudyitsroleasacontingentvariable,

whichisperhapswhyfewstudieshaveundertakensuchan

investi-gation.Nevertheless,itispossibletoavoidthisissuebyeitherusing

adifferentmeasureoforganizationaleffectiveness,orbyusinga

managerialestimateofcontrolsystemseffectiveness,despitethe

issuesraisedbyusingdatafromthesamemanagertoreportboth

onthenatureofthesystemandonitseffectiveness(particularlyif

thatpersonhasbeeninvolvedinthedesignofthesystem)

Thereare also wider issues in determining how ‘fit’ can be

assessed.ThesearewellsetoutinGerdinandGreve(2004)who

identifyanumberofdifferenttypesoffit(Cartesianvs

configura-tion;congruencevs.contingency;andmediatingvs.moderating)

eachofwhichisappropriatetodifferenttypesofhypothesis.They

alsoconsiderwhetheritistheformofarelationshipthatisbeing

hypothesized,oritsstrength.Theythenanalyzeanumberofmajor

studieswhichusestrategyasacontingentvariableandconclude

thatthereiswidespreadconfusionintheway‘fit’hasbeen

con-ceptualizedindifferentstudies(andsometimeswithinthesame

study).Thiswasfollowedbyalaterarticle(GerdinandGreve,2008)

whichattemptstosetouttheappropriatestatisticalteststoassess

whetheragiventypeoffithasbeenachievedornot

Whereseveralcontingentvariablesareusedthereare

difficul-tiesinanalyzingandunderstandingtheresultsobtainedinstudies

Apopularearlymethodwastouseaninteractioneffectina

regres-sionequation,wheretheinteractiontermisthemultiplicationof

thetwoindependentvariablesinvolved.Thisisnotproblematic

(providedtheregressionmodelisappropriatetothehypothesis

beingtested(seeabove)) butbecomes morecomplicated when

thenumber ofindependentvariables increases.Threevariables

seemstohavebeenthepracticallimit,butevenherethe

inter-pretationof a three-way interactionis complicatedand a clear

explanationoftenrequiresittobebrokendownintosub-parts,

usuallybydichotomizingonevariableintotwolevels(e.g.,highand

lowwithrespecttoitsmean)andexaminingeachsub-sampleso formedseparately,despitetheinformationlossinsuchaprocess.12

Thisprocessalsodrawsattentiontothecommonpractice,often alsoimplicitinthestatisticaltechniquesused,ofreferringtolevels

ofavariablebeinghighorlowbyreferenceonlytoothervalues

inthesampleratherthantosomeexternal benchmark,making comparabilitybetweenstudiesdifficult

Therehasalsobeenextensivedebateontheroleof appropri-atestatisticalmethodswhentestingfortheexistenceofmediating

ormoderatingvariablesincontingentrelationships.Hartmannand Moers(2003)arguedthatthereweremanycasesintheliterature wherethetheoreticalreasoningledtothepredictionofoneofthese hypothesizedrelationshipsbeingtrue,butthatthestatisticaltest usedactuallytestedtheotherhypothesis.Thisprovokedadebate

onthetopicwhichiswellsummarizedinBurkertetal.(2014)which outlinesseveraldifferentformsoffit(someofwhichhadnot pre-viouslybeendiscussed).Theyassessthebeststatisticaltechniques thatcanbeusedinvariouscircumstancesandthusprovidea use-fulguideforresearchersdevelopinglarge-scaleempiricalstudiesin thisarea.However,manyofthesetechniquesarestilldatahungry andmaynotprovefeasibleinmanycontexts,althoughstatistical techniquessuchasstructuralequationmodelingandpathanalysis canrequirelessdatatogivestatisticallysignificantresults Therearethuspracticallimitationstounderstandingthe com-plexinteractions thatmayexistbetweenmorethana veryfew contingentvariablesatatime.Itisunlikelythattherecouldever

beatimewhenwewouldbeabletopredictthebestMCSina sit-uationwherethereweremorethanasmallnumberofimportant contingentvariables,particularlywheredifferentvariablespointed

inconflictingdirections.However,thisnarrowdefinitionofwhat comprisesa‘contingencystudy’isunnecessarilyconstraining,as pointedoutbyChapman(1997),althoughhefocussesspecifically

onaccountingratherthanMCSsmoregenerally.Inanimportant sense,allresearchonthistopiccanbeseenascontingent,aswe seekto understandthe appropriate matching betweenthe use

ofanaccountingtechniqueandthecircumstancesinwhichitis used,whatevertypesofresearchmethodsareused.Chapmanalso pointsoutaconfusionintheliteraturebetweencomplexityand uncertainty,withsomescalespurportingtomeasureuncertainty actuallyfocussingoncomplexity.Hearguesforadialoguebetween traditional,quantitativeresearchapproachesincreasinglyadopted

intraditional contingency studies(evenseenas definingthem) andmorequalitative, interpretativefieldstudiesaimedat gain-inginsightintothecomplicatedprocesses bywhichaccounting

ismobilizedandusedwithinorganizations.Althoughmoreofthe lattertypeof studieshave occurredin theperiod afterhis arti-clewaspublished,themovementtowardsthestudyofMCSand performance managementsystems (ratherthanjust accounting systems)hasalsotakenplace,whichhasmadethestudies neces-sarilyevenmorecomplex,andhispleaformorefield-basedwork

isstillveryrelevant.Giventhedominanceofnon-USjournalsin reportingresearchinthisfield,itisperhapssurprisingthat qual-itativeworkhasbeensorelativelyslowindeveloping,although thereisevidencethatthisisbeginningtochange

5.6 MCS:systemsorpackages?

Therehasbeenrecentdiscussionastowhetherasetofcontrol devicesusedbyanorganizationarebetterregardedasasystem

orapackage,althoughthedistinctionbetweenthesetermsisnot alwaysmadeclear.Theconceptgoesbacktotheuseoftheterm

‘package’byOtley(1980)whereheusesittorefertotheseparate

12 This is an important area where more modern statistical techniques are proving

definition.13Thefullestdiscussionofthetopichasbeengivenby

MalmiandBrown(2008)intheirintroductiontoaSpecialIssueof

MAR.Heretheystatethatthereareseveralreasonsforconsidering

anMCSasapackage.First,thecomponentsofanMCSdonotoperate

inisolationandtheeffectofanyoneneedstobeconsideredinthe

contextoftheothercomponentsbeingusedatthesametime

Sec-ond,itseemsnecessarytoconsideranoverallpackageofcontrolsif

theuseandimpactofanewMCSelementisrelatedtothe

function-ingoftheexistingMCSpackage.Third,onlysomeMCSelementsare

accounting-basedandconsiderationneedstobegivenastohow

theserelatetobroadercontrols(suchasadministrativeorcultural

controls)andwhetherorhowthesecomplementorsubstitutefor

eachotherindifferentcontexts.Theyconcludebystatingthat“by

takingabroaderpackageapproachtothestudyofMCS,researchers

willbeabletodevelopbettertheoryoftherealimpactofinnovations

suchastheBSC,andhowtodesignMCSpackages.”(p.288)

IncludedinthesameSpecialIssueisalsoanarticlebySandelin

(2008)whichcomparesthedevelopmentofcontrolsystemsina

singleorganizationattwodifferentpointsintime.Hearguesthat,

althoughtheexternalcontextisverysimilarinthetwo‘cases’,the

controlsystemsobservedareverydifferent.Thisisused,following

Gerdin(2005a),tosupportanargumentfor‘equifinality’,namely

thefactthattwodifferentsystemsmayactuallyproduceequivalent

(orequallygood)outcomes,andthatthereisno‘onebestway’even

ifthisisconsideredincontingentterms

Malmiand Brownseemhowevertogoonly partway down

thepaththeyhaveidentified,astheypaylimitedattentiontothe

phenomenaoftenobservedinpracticethattheelementsofan

over-allcontrolsystemmaynotbewell-integratedorcoordinated.The

fullerideaofapackageshouldperhapsincorporatethisidea,which

issometimesreferredtoas‘loosecoupling’,althoughitisoftennot

clearwhetherthisisregardedasintentionallydesignedor

acci-dentlyachieved!Thisidea wasdeveloped byOrton and Weick

(1990)andMalmiandBrownsuggestthatitmightbefruitfulto

followasimilarapproachincharacterizingthelinkagesbetween

theelementsinanMCSpackage

ThemaincontributionoftheMalmiandBrownarticleseems

rathertobethetypologytheygiveofthevarious categoriesof

control that forman overall package These are seenas falling

intofivedifferentcategories,mostcentrallyPlanningand

Cyber-neticControls(essentialexpostandexantecontrol),togetherwith

RewardandCompensation.Thesearesupplementedby

Adminis-trativeControls(whichincludebothgovernanceandorganizational

structure)andallareincludedwithinawidersetofCultural

Con-trols.Thisprovidesavaluablesupplementtothetwelvequestions

suggestedbyFerreiraandOtley(2009)intheirproposed

frame-workforthestudyofoverallcontrolsystems,althoughtheymake

littlereferencetotheideaof packagesexceptin theirquestion

aboutcoordinationbetweencontrols

AlimitedconceptualizationofanMCSanditscomponents

per-meatestheliterature.Mostarticlesselectjustonecomponentof

anoverallsystemforstudy(suchasbudgeting,useofaBSC,or

compensationsystem)andproceedwithoutanyreferencetothe

othercomponentsthatsurroundit.Itisrareforanattempttobe

madetocapturethetotalityofanoverallsystem,althoughsome

oftheusersoftheFerreiraandOtley(2009)frameworkhavetried

tomoveinthisdirection.Thisisgratifyingtomeastheframework

wasputforwardinordertomakeitmorefeasibleforresearchers

toattemptsuchatask.Someusehasalsobeguntotakeplaceofthe

13 My recollection is that I felt unhappy with using the term ‘system’ because it

seemed to imply a designed system of well-coordinated parts, and many overall

MCSs did not seem to possess this property I used the term ‘package’ to imply a set

MalmiandBrown(2008)categorization,butthesestudies com-prisejustasmallpartofextantwork.GrabnerandMoers(2013) havealsotackledtheproblemofmorepreciselydefiningtheidea

ofinternalconsistencywhichhasbeenusedtodefineanoptimal (perhapscloselycoupled) MCSpackage However,theirworkis primarilyanalyticandismainlyconcernedwiththe conceptual-izationof‘internalconsistency’basedonanumberofassumptions, beforeaddressingsomeimplicationsforempiricalstudies.There

isthereforescopeforthisissuetobeexaminedfromanempirical perspectivetocomplementtheiranalyticapproach.Oneexample

ofsuchworkisGongandFerreira(2014)whichseeksto exam-inewhetherconsistencyinmanagementcontrolsystemsdesign choicesaffectsfirmperformance,withtheirresultsshowinga pos-itiverelationship

Theimportanceofoverallcontexthadalsobeenemphasized

byNixonandBurns(2005)intheirintroductiontoanearlier Spe-cialIssueofMAR.Theydrewparticularattentiontotheideathat changehaditselfchangedbybecomingmuchmorerapid.Theidea thatperiodsofchangearepunctuatedbyperiodsofequilibrium hasbecomeoutdated.Theyalsodrewattentiontothecontinued existenceoftwogaps:first,thatbetweenextantmanagement con-trolliteraturesandmanagementpractice,andsecond,betweenthe managementcontrolliteratureandotherliteratures.Theyargue thattheseissues“increasethepersuasivenessoflongstanding criti-cismsofthetheoreticalfoundationsofmanagementcontrol”(p.262) ThisviewisstronglyreflectedinonearticleinthatSpecialIssue

by Collier (2005) which providesan interesting perspective on several control issues, namely thechanges that occur in man-agementcontrolover anorganizationallifecycle,thedegreeof consistencyamongdifferentcontrols,andtheinteractionbetween formal,systems-basedcontrolsandinformalcontrols.Inaddition,

hearguesthatboththeframeworksofFerreiraandOtley(2009) andSimons(1995)paytoolittleattentiontobeliefscontrols,and concludes thatmodernorganizationswill havetodevelop con-trolswhichallowthembothtocompetetodayandtopreparefor tomorrow

Finally, Mundy(2015) hasdeveloped the ideaof loose cou-plinginMCSsfollowingearlierworkbyDemartini(2011).Although developedin bothorganizationtheoryandeconomics,thisidea hasyettobefullyappliedtoMCSs.Mundytakestheviewthat MCSswithinmanyorganizationsconstituteapackageofdistinct managementcontrolelementsthathavebeenseparatelydesigned andimplementedwithoutanoverallintentionorcoordination,but whereeachelementaimstofacilitatetheattainmentofdifferent aspectsoforganizationalgoals.Sheappliesthistoexaminethe pat-terns ofinteraction observedwithina MCSpackage in a single organization,and interpretstheseasanexemplarof loose cou-pling.Ofparticularinterestisherdocumentationofmanagement

‘workarounds’,thatis,informalpracticesthatenabletheoverall packageofcontroltofunctionbetterthanobservationofjustthe formalsystemsmightimply.Shecharacterizesthecouplednature

ofinterrelationswithinanMCSpackageusingfourdimensions: mode,directness,directionality,andfrequency,anddemonstrates theactualfunctionalityof aspecificpackageofcontrolsdespite itspartshavingbeenimplementedatdifferenttimes,bydifferent individualsandfordifferentpurposes

ThesestudiesindicateanemergingviewthatoverallMCSsare influencedbytheircontentandtrajectoryofdevelopmentaswellas

bytraditionalcontingentfactors.Asexternalchangeoccurs,control systemsareadapted,oftenbytheadditionofadditionalfeaturesor systems(andperhapslessbytheremovalofpracticeswhichhave becomeoutdated,asthesetendtofallintodisuseratherthanbeing removed).Ratherthanmovingfromoneequilibriumtoanother, suchsystemsresemblepackageswhichareconstantlybeing modi-fiedanddevelopedwiththeever-presentdangerthattheybecome internallyinconsistentandincoherent.Althoughthiscanbe(and

practiceMCSpackageswillinevitablycontaininconsistentand

con-tradictoryelements.Itappearsthatinformalprocessesarelikely

toplay anincreasinglyimportant roleinhelping preservetheir

continuingfunctionality

6 A characterization of MCSs and their context

Theabovecritiqueshaveconcentratedonthespecificationof

differentcontingent(independent)variables thathavebeen

fre-quentlyusedintheliterature.Itissurprisingthatasimilaramount

ofattentionhasnotbeendevotedtothedependentvariableofthe

MCSandaspectsofitsdesignanduse.Bycontrast,thishasgenerally

beentreatedinafragmentarymannerwithjustoneortwoaspects

selectedbyeachreportedstudy,withverylittleworkattemptingto

gainaholisticviewoftheoverallsystemsinusebyanorganization

Admittedly,thisisasignificanttasktoattempt,particularlywith

thelackofframeworkstoassistincategorizingsuchMCS

compo-nents.Forexample,FerreiraandOtley(2009)focusprimarilyon

thepurposesservedbythecomponentsofsuchsystems,whereas

MalmiandBrown(2008)attemptabasicclassificationoftoolsand

techniques,butagainconcentrateontheiruse(e.g.,planning,

con-troletc.).Itseemslikelythat moreusefulprogresswillnowbe

gainedbyagreaterfocusontheMCSitself,howitselementsmay

mostusefullybeconceptualized,andhowtheyareinter-related

Thisrequiresaricherconceptionoftherealityoftheactualcontrol

systemsextantinorganizations,andanattempttocharacterizethis

issetoutbelow

Organizations exist in an environment of considerable

uncertainty.14 Business organizations, in particular, working

inacompetitiveenvironmentcannevertotallypredicttheactions

oftheircompetitorsorwhattechnologicalinnovationmayoccur

andonwhattimescale.Theyalsodependuponother

organiza-tionsaseithersuppliers,customers,serviceprovidersornetwork

partners,andtheserelationshipsaresubjecttoongoingchange

Theoveralleconomicenvironmentisnowglobalandincreasingly

unpredictableasnosectorisimmune tochangeshappeningfar

away.Eventhepublicsectorissubjecttosomeofthesefeatures

andalsobythevagariesofpublicpolicy,changesin whichcan

sometimesbequitesudden and unpredictable.In addition,the

internalorganizationalenvironmentisalsosubjecttoconsiderable

uncertaintiesabouthowemployeeswillcommunicate,coordinate

andgenerallybehaveinwaysconducivetoefficientoperations.15

Ofcourse,differenttypesandamountsofuncertaintyexistin

differentenvironments,andthesedifferenceshavebeenthefocus

ofonemajorcontingentvariable.Butitneedstoberecognizedthat

substantialuncertaintyispresentinallorganizationsandthatthey

needtocopewiththis,asarguedinOtley(2014).ThusMCSsneedto

beabletooperateinsuchanuncertainenvironmentandmayneed

tobeadaptedtocopewiththisevermorecommoncircumstance

Forexample,theBeyondBudgetingmovementhasclaimedthat

inthis‘new’environmentbudgetingisfundamentallypastits

sell-bydateandneedstobereplacedwithabettersystem(Hopeand

Fraser,2003),althoughitnowseemstohavemodifieditsclaimsof

beinganalternativesystemtothemoremodestsuggestionthatit

comprisesasetofusefultools(seeHansenetal.,2003)

Neverthe-lessthemajorityoforganizationsofanysizestillhavebudgetary

14 See Otley and Soin (2014) for a set of essays on how uncertainty affects the

design and operation of MCSs.

15 I well remember as a PhD student being ‘reprimanded’ by my teacher of

organi-zational behaviour for concentrating excessively on dysfunctional behaviour “Isn’t

what is really surprising, he suggested, is that organizations actually work as well

as they do, and shouldn’t we concentrate on trying to understand how they achieve

systems,althoughthesemayhavebeenadaptedinavarietyofways (e.g.,morefrequentupdating;shortertime-spans;rollingbudget systemsetc.;differentpatternsofbudgetusefordifferentpurposes etc.).Whatisevidentisthatbudgetarycontrolhasbeendisplaced fromthe central place it occupiedforty years ago and supple-mentedbyothersystems,mostnotablyBalancedScorecard-type systemscontainingavarietyofnon-financialmeasuresof perfor-mance.Inaddition,theseperformancemeasurementsystemsare oftencomplementedbyperformance-relatedpaysystemswhich relyonfurther(andsometimesdifferent)performancemeasures, andwhichareusuallydifferentatsuccessivehierarchicallevels

Acentralfeatureoftheelements(sub-systems)ofanoverall MCSisthattheycompriseoflargelyindependentsystemsintheir ownright.Forexample,westillhavebudgetarycontrolsystems (althoughtheways inwhich theseareusedmayhavechanged markedly),butalsonon-financialperformancemeasures,perhaps integratedintoadashboardorscorecard,riskmanagement sys-tems,incentivepaymentsystems,corporategovernancesystems andstrategiccontrolsystemsaswellasmanyothers.Thesehave generallybeendevelopedbydifferent(groupsof)peopleat differ-enttimes,oftenwithaparticularandlimitedpurposeinmindat thetimeoftheirimplementation.Inaddition,theymayimpinge differentlywithin eachorganizational function,and atdifferent hierarchicallevels.Thusitoftenseemstomakemoresensetothink

ofanorganizationalcontrol‘package’asthedifferentelementsare likelynottobeperfectlycoordinatedwitheachother.For exam-ple,theperformancemeasuresdrivingaseniormanagementbonus schememaynotbethose specifiedin thetop-levelBSC(areal example!)

This process is dynamic Changes may be made from time

to time, often with the purpose of aligning disparate systems with each other These will often be successful in their own right,althoughthereareoftensignificantlagsinalignmentbeing achieved (payment systems being particularly ‘sticky’ in this respectastheyoftenrequirerenegotiationwithgroupsof employ-ees;andcomputersystemsareinvariablyslowtogetupdated) However,othernewsystemsmayalsobeintroducedthathave theirowndifferencesfromtheextantpackage,andthespeedof improvementoftenlagstherateofintroductionofnewelements Theways inwhich employeesusetheelementsavailableto them oftenchanges over time, sometimesquiterapidly When Mundy(2015)documentedthe‘workarounds’thatwereadoptedin theorganizationsheobserved,thesecouldbegenerallydescribed

asfunctionalinthat theyallowedill-coordinatedsystemstobe madeworkablebytheseinformalchanges.Additionally,in multi-nationalorganizationsanotherlayerofcomplexityoftenarisesin thatsystemswhichoperatewellinonenationalenvironmentmay workvery differently in anothernational culture Mergers and acquisitionsalsooftendisplayinstancesofsystemschangewhich provesineffectivebecauseofthedifferingorganizationalcultures

ofthecombiningorganizations

Finally,moreorganizationshavebecomepartofavaluechain thatcoversseveralindependentorganizations,whichhaveno com-monownershiporsuperiorauthority,yetstillneedtocoordinate theiroperations.Manyoftheseseemtohaveoneorganizationin

adominantposition,eitherbecauseitissignificantlylargerthan theothers,orithasamonopolyovertheroutetotheultimate con-sumer,andhereitssystemswilltendtobeimposed.Forexample, mostcasesofopenbookaccountingseemtodisplaythis charac-teristic.Butothersarecomprisedofmoreequalpartnerswhoneed

todevisearrangementstoachievebettercoordinationandcontrol, andareofteninacontinuingprocesstoadapttheirarrangements

Ithasbecomeusualforsuchpartlyco-ordinatedcombinations

of elements into an overall system to be labelled as ‘loosely-coupled’althoughthistermseemstocoverbothintentionaldesign andaccidentaloutcomes.Itseemsprobablethatmostofthe