Towards a better modelling and assessment of construction risk:Insights from a literature reviewAbdulmaten Taroun Lecturer Business Systems (Project Management), Department of Management and Business Systems, The University of Bedfordshire Business School,Polhill Avenue, Bedford, UKReceived 24 July 2012; received in revised form 30 December 2012; accepted 19 March 2013Available online xxxxAbstractThis paper reviews the literature of construction risk modelling and assessment. It also reviews the real practice of risk assessment. The reviewresulted in signicant results, summarised as follows. There has been a major shift in risk perception from an estimation variance into a projectattribute. AlthoughtheProbabilityImpact riskmodel isprevailing, substantial effortsarebeingput toimprovingit reectingtheincreasingcomplexity of construction projects. The literature lacks a comprehensive assessment approach capable of capturing risk impact on different projectobjectives. Obtainingarealisticproject risklevel demandsaneffectivemechanismforaggregatingindividual riskassessments. Thevariousassessment tools suffer from low take-up; professionals typically rely on their experience. It is concluded that a simple analytical tool that uses riskcost as a common scale and utilises professional experience could be a viable option to facilitate closing the gap between theory and practice of riskassessment. 2013 Elsevier Ltd. APM and IPMA. All rights reserved.Keywords: Construction; Literature review; Project management; Risk assessment; Risk assessment practice; Risk cost; Risk modelling1. IntroductionNo construction project is risk free. Risk can be managed,minimized, shared, transferred or accepted. It cannot be ignored(Latham, 1994, p.14). Theconstructionindustryisoftencon-sidered as a risky business due to its complexity and the strategicnatureofitsproducts.Itinvolvesnumerousstakeholders,longproductiondurationandanopenproductionsystem, entailingsignificant interactionbetweeninternal andexternal environ-ments(BSI-6079-4,2006).Suchorganisational andtechnolog-ical complexitygenerates enormous risks (Zouet al., 2007).Unfortunately, the construction industry has a poor reputation inrisk analysis when compared with other industries such as financeor insurance (Laryea, 2008). Although every step of a risk man-agementprocesshasreceivedhugeattentionfromresearchers,it seems that risk assessment is a controversial issue (Baloi andPrice, 2003). Traditionallythefocushasbeenonquantitativeriskassessment (TahandCarr, 2001) despitethedifficultiesencountered in obtaining objective probabilities, and frequencies,in the construction industry. This difficulty stems from the factthat construction projects are very often one-off enterprises(Flanagan and Norman, 1993). This reality is a key driver behindtheobligationofproject managersbeingtorelyonsubjectiveprobabilitiesasWinch(2003)concluded.Infact,riskinmanycases is subjectively dealt with through adding an approximatecontingencysum(Kangari andRiggs, 1989). Therefore, indi-vidual knowledge, experience, intuitive judgement and rules ofthumb should be structured to facilitate risk assessment (Dikmenet al., 2007b).Risk assessment is inherently related to risk modelling. TheProbabilityImpact (PI)riskmodel isprevailingandriskisusually assessed through assessing its probability of occurrenceand impact. However, the PI risk model was subject to criticismfromresearcherswhodiscussedpotential improvementsinit.Moreover, researchers have investigated different theories, toolsTel.: +44 1234793141.E-mail address: abdulmaten.taroun@beds.ac.uk.www.elsevier.com/locate/ijproman0263-7863/$36.00 2013 Elsevier Ltd. APM and IPMA. All rights reserved.http://dx.doi.org/10.1016/j.ijproman.2013.03.004Please cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights froma literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004Available online at www.sciencedirect.comInternational Journal of Project Management xx (2013) xxxxxxJPMA-01518; No of Pages 15andtechniques foraiding risk assessment.Unfortunately, thereis a clear gap between the theory and practice of risk modellingand assessment. Hence, it is of crucial importance to understandthe actual practice of risk analysis and review the developmentof construction risk modelling andassessment in an attempt toresearchviablealternativesthat maycontributetoclosingthisgap. This paper presents the result of an extensive review of thepublishedliteratureofconstructionproject riskmodellingandassessment inEnglish. Theresearchfocusedmainlyonpeer-reviewedarticlespublishedinacademicjournalsspecialisedinconstructionmanagement, project management, riskanalysis,and management science. The following databases were utilisedfor researching relevant papers: Science Direct, Web of Science,ABI-Inform(Proquest), Business Source Premier (EBSCO),Emerald, andSageManagement &OrganizationStudies. Tofurther support reviewingthepublishedliterature, theGoogleScholar search engine was deployed. The following key wordswere usedinthe search: project risk, constructionrisk, riskanalysis, risk assessment, risk modelling, and risk management.These words were usedafter discussingthemwithresearchcolleagues. However, different combinations of them were usedtovalidatetheextensivenessof thesearchresults. Moreover,other key words were tried to investigate any differences in thesearchresultssuchas: riskmodel, riskmodelling, uncertaintyanalysis,andprojectuncertainty.Besidescomputerisedsearch,manual bibliographic search was also used. In many cases,reviewingthepapershelpedinidentifyingrelatedpapers. ThereviewprocesstookplacebetweenOctober2008andAugust2009. However, regularsearchactivitieshavebeenconductedsince then to keep the reviewresults updated. The search targetedall of the available articles in the databases in order to review thehistorical development of risk modelling and assessment. Hence,there was no time restriction when searching the databases. Asa result, around 400 articles were reviewed. Eventually, 68 oneswere considered as most relevant to the research aim and weresubject to a detailed review. These papers, detailed in Appendix(A),coverthelastthreedecadesofprojectriskmodellingandassessment history. Tobe included in thefinallist,papers hadtomeet thefollowingcriteria: 1) provideamethodologyforassessingproject risk; 2) usespecifictheoryor techniqueforassessingrisk; 3) present anattempt toimproveproject riskmodelling; and 4) relate to construction or project managementdomain. In the next section of the paper, a chronological reviewof these 68 papers is provided. Later, the paper analyses and theresults of the literature reviewwill be discussed to enable elicitingthe main themes and developmental trends. After that, a reviewoftheactual practiceofriskanalysisisdisplayedinordertocomplement the review results and to enable defining the char-acteristics of practical alternatives. The paper ends with a sum-mary of the key findings and conclusions.2. Literature reviewRisk analysisin construction industryis not new.It has itsroots sincethedevelopment of theProgramEvaluationandReview Technique (PERT) in the 1950s for tackling uncertaintyin project duration. Conventionally, risk has been dealt with as anestimation variance benefiting fromthe dominance of ProbabilityTheory (PT). In the 1980s, however, risk began to be perceived asa project attribute and Risk Management (RM) became a well-established project management function. During the 1990s re-searchers investigated different theories to account for the specialnatureofconstructionrisk,andafterthebeginningofthenewmillennium risk assessment flourished as a hot research topic.2.1. Before the 1980sAlthoughthe origins ofrisk analysiscanbe tracedback toas far as 3200 BC (Baker et al., 1999b), risk had not appearedinconstructionliterature until 1960s (Edwards andBowen,1998). Baker et al. (1999b) argued that the term risk analysiswasusedfor thefirst timebyHertz(1964) whoutilisedthecomputer for generating probability distributions of investmentprojectsratesofreturn. Reviewingliteraturerevealsthat riskanalysis publications started in the USAwhere risk was consideredimplicitly when researching other problems like bidding and costanddurationestimation. Riskwas modelledas anestimationvariance and RM was perceived as a way of reaching more accu-rate estimates during the tendering stage. According to Edwardsand Bowen (1998), statistical methods were initially used beforeemployingMonte-CarloSimulation(MCS) duringthe 1970s.Despite the dominance of the probabilistic methods and MCS, thedearth in risk analysis publication is very evident in that era; veryfew articles about risk analysis can be referred to like Carr (1977),Friedman(1956), Gates(1960, 1967, 1971), GatesandScarpa(1974), Morin and Clough (1969), and Spooner (1974). Regardingrisk management, it was the end of the 1970s when project RMstarted to became an essential component of project management(MernaandAl-Thani, 2008). Actually, reviewingthelitera-ture reveals that the beginning of the 1980s is the actual startof perceiving RM as an independent project management func-tion and research domain.2.2. The 1980sInthe1980sPT-basedtoolsandMCScontinuedtodomi-nate risk assessment. However, Fuzzy SetsTheory (FST) wasintroduced at the end of this decade as a viable alternative fortackling subjectivity in construction risk assessment. ChapmanandCooper (1983) presentedoneof theearliest attempts tostructure project risks and, systematically, identify their sources.Theyintroducedtheriskengineeringapproach, whichinte-grateddifferent toolsandtechniqueslikePERTanddecisiontrees, forcombiningrisk events andproducingprobabilitydis-tributionsof activitiesandproject durations. Hence, riskwasmodelledas a distributionvariance of anactivityor projectduration. Diekmann(1983), however, modelledriskasavari-ationof cost estimation. Herevieweddifferent toolsusedforproducing a probabilistic estimate of project cost and used MCSfor such a purpose. Contrary to the previous two papers, Barnes(1983) modelled risk as probabilityandimpact (PI) with riskimpactdefinedasavarianceincostestimate. Inasubsequentpaper, Cooper et al. (1985) presentedamethodfor assessingproject cost risk. Ahierarchical riskbreakdownstructurewas2 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights from a literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004developed with project cost risk at the top of the hierarchy. Riskwas modelledagainas avariationof cost estimate. Beeston(1986)alsomodelledriskasanestimationvarianceandusedMCS to generate more accurate estimates. Clark and Chapman(1987) also dealt with risk as an estimation variance while dis-cussing the risk analysis methodology developed by the ProjectsDepartment of BP's Group Engineering and Technical Centre.In one of the very few papers that highlighted the need forconsideringtheimpact of riskonspecificproject objectives,Franke (1987) advocated assessing risk impact financially. Heproposed the use of risk cost as a common scale and adoptedthePI riskmodel. Here, riskcost isrecommended, for thefirsttime,tobeusedformeasuringthedifferenttypesofriskimpact on project success factors. Such an approach provides adetailed and comprehensive risk impact assessment. Moreover,it isapractical wayof quantifyingriskimpact andaneasyapproachforincludingriskimpactinproject cost estimation.Althoughthis paper proposes apioneeringapproachfor as-sessing risk impact, the overall project risk level is treated in arather simplistic manner. Project risk level is obtained throughsumming up the individual risk costs ignoring any interdepen-dencies between these risks. Such an aggregation process mayresultinanunrealisticproject risk.Inthelate1980s,KangariandRiggs (1989) discussedthe potential usage of FSTforriskassessment. Itistheearliest paperthatrecommendsFSTfor handlingthesubjectivityof constructionriskassessment.Indeed, itisaveryimportantpaperandoneofmostcited. Itpresents an objective evaluation of the merits and shortcomingsof FST for assessing construction risk. The concept of FST wasdeveloped by Lotfi Zadeh in 1965 to represent the uncertainty,usingnaturalwords, whichthelackofprecisecrispnumberscancause. FSTrepresentsuncertaintythroughlinguisticvari-ablesandmembershipfunctions. Everylinguisticvariableisrepresented by a fuzzy set, a set of numbers, associated with amembershipfunction. The membershipfunctionis a set ofnumbers between 0 and 1 representing the degrees of member-ship with 1 standing for complete belonging and 0 standing forcomplete non-belonging. Hence, the partial belonging to a set isthe key difference between FST and the crisp set theory whichis based on complete belonging or complete non-belonging to aset (Liu et al., 2002). As will be illustrated in the next sections,FSThas flourished over the last two decades in assessingconstruction risk.2.3. The 1990sDuringthe1990sconstructionriskmodellingandassess-ment gained momentumand became a hot research topic.Researchers primarily used two theories: PT and FST. Yet, inparticular, theywere keentoresearchother tools andtech-niques like the Analytical Hierarchy Process (AHP). Hull (1990)introduced different models, based on MCS and PERT, to assessproposalrisksfromcostanddurationperspectives, whileYeo(1990) presenteda contingencyengineering method, usingboth a range estimate method and PERT, to assess project costriskandtoestimatecontingencysums.Itisoneoftheearliestattempts to estimate risk contingency in a systematic manner. AlBahar and Crandall (1990), in turn, used influence diagrammingandMCS to assess risk.They adoptedthePIrisk modelandprovidedasystematicapproachfor identifying, assessingandmanaging construction risks.Pioneering its application in construction, Mustafa andAl-Bahar (1991) adopted the AHP to assess constructionproject risk. Becomingoneof the most citedpapers intheliterature; it applied the concept of value and weight of AHP toassess risk probability and impact. It is evident from reviewingthe literature that AHP has received a paramount attention frommanyresearches. AHPwasdevelopedbyThomasSaatyandpresented in Saaty(1980).It has receivedworldwiderecogni-tionasapowerful Multi CriteriaDecisionMaking(MCDM)tool. It isaveryuseful tool forstructuringcomplexdecisionmaking systematically. One of the core attributes of AHP is itsability to facilitate the decision maker's personal judgement inquantifying relative priorities of decision alternatives on a ratioscale. There is no doubt that AHP provided a vital approach forstructuringtheincreasingcomplexityofconstructionprojectsand construction risk assessment. In fact, Mustafa and Al-Bahar(1991) presenteda verygoodevaluationof the merits andlimitations of AHP and its suitability for assessing constructionrisk.Influence diagramming appeared again in Diekmann (1992)representing risky situations and MCS and FST were deployed todo risk assessment calculations. Similarly, Huseby and Skogen(1992) usedinfluencediagrammingandMCStoaccount fordependencies between risks and then assess them. They devisedsoftwarecalledDynRisktofacilitateusinginfluencediagram-ming and MCS. In one of the few attempts to price risk system-atically, Paeket al. (1993) usedFSTtoassist contractors indeciding on bid prices. Likewise, Tah et al. (1993) used FST toassist contractorsinestimatingriskcontingency; thePI riskmodel was adopted. AHP appeared again in Dey et al. (1994) inariskassessmentmethodologywhichcombinedobjectiveandsubjective assessments; thePIrisk modelwasagainadopted.Riggset al. (1994)proposedanapproachforquantifyingandintegratingtechnical, cost, andschedule risks throughutilityfunctions. AHP was used to elicit the utility functions and utilitywas usedasacommonscalefor assessingtheattractivenessofdifferentscenarios.Actually, theproposedmodelcouldnotassessrisk; it couldonlyassesstheutilitiesofdifferent riskyscenarios. Williams (1995) conductedanextensive literaturereview of the available tools and methodologies of constructionRM. Hefoundthatriskassessmentusedtofocusoncostanddurationrelatedrisks; qualityrelatedriskswereneglected. Hereportedthelackofresearchtowardsassessingriskimpactondifferent project objectives simultaneouslyandattributedthatto the lack of a common assessment scale. Like Franke (1987),he recommended risk cost as a feasible solution. AHP was usedagainbyZhi (1995)toassesstherisklevelsofoverseascon-structionprojects; thePI model wasadoptedandAHPwasdeployed with minor modification; the impact assessments fell ina [01] range instead of the AHP's formal 19 ordinal scale.Thelimitations of the PI riskmodel werediscussedinWilliams(1996). Hearguedthat multiplyingprobabilityandimpact produced an expected value of risk which is misleading3 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights froma literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004and cannot be simply used to prioritise project risks. As a result,hesuggestedconsideringbothprobabilityandimpactforpri-oritising risks. He also concluded that a three dimensionalriskmodel: ProbabilityImpactPredictability, recommendedby Charette (1989), was a viable alternative to the PI model.Similarly, Ward(1999) discussedthelimitations of thePImodel andurgedfor improvingriskassessment. Heinsistedthat the multiple impacts of risk on specific project objectivesshould be considered before calculating the overall risk impact.Besides, he criticisedthe methodof generatingproject risklevel throughsummingupthenumerical scores of separateProbabilityImpact grids representing the assessments of proj-ect risks. As an alternative, he proposed using a weighted sumof alphabetical ratings. Although the alphabetical rating may beafeasiblealternative, thisapproachmaynot beeasilyappli-cable for assessing project risk level. For instance, a risk ratingcouldbeexpressed,basedonitsimpactsonthreeprojectob-jectives, as3A + 2C + 5D. Sucharatingisverydifficult touse when aggregating a large number of risk assessments withdifferent scores on various project objectives.Wirbaet al. (1996) usedFSTtoassessrisklikelihoodofoccurrence using linguistic variables. In this paper, risk impactwasconsideredasthecostofariskresponsestrategy. Whilethispaperiswidelycited, thereisaconcernabout usingthefuzzy weighted mean method aggregating risk assessments; thisisakeypoint of weakness inFSTasit onlycalculatestheweighted average of the individual assessments. Although FSTbecame more common, PT-based methods continued to be used.Dawood(1998) usedMCStoestimateanactivityor projectduration with risk modelled as an estimation variance. Mulhollandand Christian (1999), however, used PERT to estimate a projectduration. Again, the variance of project duration representedproject schedule risk; the larger the variance, the greater the riskassociated with project duration.2.4. The new millenniumSince 2000, attempts at better modelling and assessment ofconstruction risk have intensified and tools have become moresophisticatedbenefitingfromtheavailabilityofhighcapacityPCs. As a result, risk assessment, very often, was facilitated bydecision support systems (DSSs). Despite their limitations,AHPandFSTbecametheprincipal approachesforhandlingill-definedandcomplexproblemswithsubjectivityinvolved.Yet, PT-based techniques continued to appear in literature, butwith less frequency if compared to the previous eras.Chapman and Ward (2000) criticised the use of ProbabilityImpactgridtosizeriskarguingthatitgeneratedunnecessaryuncertainty by over-simplifying the estimates of risk probabilityandimpact. As analternative, theyproposedtheminimalistapproach which identifies risks and assesses their probabilities andimpacts by specifying ranges instead of single scores. Hastak andShaked (2000) deployed AHP for assessing risks in internationalconstruction projects and adopted the PI risk model. Althoughthedevelopedmethodprovidedanassessment of project risklevel, the assessment methodologywas over-simplistic. Riskswere subjectively assessed using a predetermined scale of 0100,where 0 implies no risk and 100 implies maximumrisk and projectrisklevel was definedas theweightedsumof theindividualassessments.TahandCarr(2000)assessedriskprobabilityandimpact and risk interdependencies using FST. They tried toovercome the limitation of the fuzzy averaging rule by introducinga new aggregation formula based on the maximum risk estimate,Emax, using a modification factor () as follows: E = Emax.Althoughthisaggregationruleisavital alternativetotheaver-aging one, it might not be an appropriate choice for every case. Forinstance, one may struggle in using it in the case of having morethan one predominant risk factor. In fact, the researchers agreedthat their method needed further investigation.The methodology of the Department of Contract andManagement ServicesinWesternAustraliausedfor rankingprojects based on their risk levels was adopted by Baccarini andArcher (2001) for assessing risk. The methodology utilises thePIriskmodelandcalculatesariskscoreforeachofprojectcost, time or quality. Although the methodology considers riskimpact onspecificproject objectives, theaggregationruleisquestionable. Thescoresofrisklikelihoodofoccurrenceandimpacts on project cost, time and quality are averaged and thenmultipliedfor generatingindividual riskscores. Eventually,project risklevel isdefinedasthehighest oftheriskscores.Suchan aggregationruleis over-simplisticandmaynot yieldrealisticassessments. Ben-DavidandRaz(2001) presentedamodel that optimises risk reduction actions through generatingthemost cost-effectivecombinationofriskreductionactions.The PI risk model was adopted and the impact was measuredinmonetaryterms.Themodelallowsincludingtheimpactofpositiveriskimpacts suchas potential savings or additionalprofit. Hence, the expected risk impact, positive or negative, istheproduct of theriskprobabilityandimpact. Althoughthemodel providesapractical methodfor gaugingproject risks,the overall project risk level, or the total project risk exposure,is over-simplistic. It is defined as the sum of the expected riskimpacts whichimplies anassumptionof riskindependencewhich generates unrealistic project risk level.Dey (2001) proposed a DSS for managing risk in the earlystages of a construction project using AHP and decision trees.Thetoolaimstoidentifythebeststrategyformanagingcon-struction project risk through the expected monetary value(EMV)ofeveryriskresponsestrategy. Theapproach, hence,does not quantify the impact of any risk; it recommends the riskresponsescenariowhichhasthelowestexpectedcost. Inthispaper, riskimpact wasrepresentedbytheextracost andtheextratimewhichwascalculatedtomonetaryequivalent. TheDSS considered risk impact on project cost and duration only; itdidnot consider riskimpact ontheother project objectives.Moreover, it did not suggest a specific way of aggregating thevariousriskimpacts. ThePT-basedtoolsappearedagain; XuandTiong(2001) usedstochasticprogrammingasatool forreaching accurate pricing. As in similar cases, risk was modelledas a variation in cost estimation. Tah and Carr (2001) utilised theFST for assessing risk and providing project risk rating. Althoughthey calculated separate ratings of risk impact on project duration,cost, quality and safety, they did not propose a methodology forcombining these separate ratings. Patterson and Neailey (2002),4 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights from a literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004in turn, used linguistic variables for assessing risk probability andimpact. However, theyproposedcalculatingproject risklevelthrough averaging the individual risk assessments.It seemsthat researchershavetriedeverypossibletheoryandtechniqueforassessingrisk. Moreover, aggregatingindi-vidualriskswasalwaysakeytask;researcherstrieddifferentapproachesfor generatingarepresentativeproject risklevel.Baloi andPrice(2003)presentedaveryimportant reviewofthe available tools and methodologies for risk assessment. Theyconcluded that FST was a key alternative to suit the case of theconstruction industry.The nature of risk and the meaning of the term risk was amatter of concern for the researchers. Ward and Chapman(2003) proposed using the term project uncertainty managementinsteadof projectrisk management.They argued that thetermriskwasconveyingamessageofthreatwhereasthewordun-certainty was more suitable for saying that there was opportunityas well. A different meaning of the term risk was used by Jannadiand Almishari (2003). They defined risk as the potential damagethat mayaffect personnel or property. Hence, riskwas spe-cificallyusedtostandforhealthandsafetyproblems. Jannadiand Almishari (2003) proposed a three dimensional risk modelcontainingriskprobabilityof occurrence, severityof impactand exposure to hazards. They also devised software called RiskAssessor Model (RAM) to generate risk scores. However, theydidnot propose aspecificmethodologyfor aggregatingriskratings. The different natures of risk were also addressed by Choiet al. (2004) who deployed FST for analysing risk in the under-groundconstructionprojects. TheydevisedanintegratedDSScapableof handlingdifferent typesof uncertainty, usingbothfrequenciesandsubjectivejudgements, basedontheavailableamount of information.2.5. Post 2005Thesharpincreaseinthenumber of riskassessment andmodellingpaperspublishedafter 2005isremarkable. More-over, anevident trendfor improvingriskmodellingcanbedetected. The majority of the published papers dealt with risk asaproject attributerather thananestimationvariance. Suchamajorshift inriskperceptionhasresulted, inmanycases, inintegrating risk assessment in comprehensive decision makingframeworks. The increasing complexity of risk assessment wasmanifested in the proposed development in risk modelling andthe extensive deployment of sophisticated DSSs.Shang et al. (2005) used FST to assess risk probability andimpact and developed a DSS for aiding construction riskassessment in the design stage. The DSS allows different projectmembers to access via the WWW and to express their assess-ments. These assessments are then weighted and synthesised. Anew risk model was introduced by Cervone (2006) to considerthe interdependencies between project risks. It was argued thatthePIriskmodelsuffersfromtheassumptionthatrisksareindependent from their environment, which is not the case in aproject context. Theproposedmodel includedanadditionaldimension called risk discrimination using the definitionofKendrick(2003).AccordingtoCervone(2006,p.260)riskdiscrimination is designed to gauge the impact of a risk on theoverallframeworkoftheproject,ratherthanlookingateachrisk as anindependent variable withinthe project. Intheproposedmodel,riskismodelledasfollows:R = (P I) / D.The new risk model appreciates the interdependencies betweenrisks through reducing their independent scores P I. Thereduction is performed through dividing the independent P Iscore by the risk discrimination factor D. In fact, the same modelwas adopted byNieto-Morote andRuz-Vila(2011)as willbeillustrated later. Risk interdependency issue was also addressedby Poh and Tah (2006). They used influence networks tocapture interdependencies among the factors affecting the dura-tion and cost of a construction activity. However, they did notsuggest a tool for assessing cost and duration risk. In a differentattempttomodeltheinterdependenciesbetweenprojectrisks,Thomas et al. (2006) used fault tree to model different scenariosandutilisedlinguisticvariablestoassessriskprobabilityandimpact. Theyattemptedtoimproveriskassessmentsbycon-sidering the opinions of different experts; they called this methodFuzzy-Delphi. The proposed model does not assess project risk;instead it provides a tool for assessing the risk levels of specificrisk scenarios.DikmenandBirgonul (2006)usedAHPwithinaMCDMframeworkforassessingriskandopportunityininternationalconstructionprojects. ThePI riskmodel was adoptedandthe overall risk level of a project was calculated by summing upthe individual risk assessments. Truly, the rather simplisticapproach of generating project risk level is questionable.Moreover, the model cannot be used in a straightforward wayto quantify or assess project risk; itcompares the risks of oneproject withtheircounterpartsinotherprojectsandprovidesrelativeriskscores. Infact, therelativenatureof theresultsgeneratedbyAHP-basedmodelsisoneofthekeylimitationsof usingAHPfor riskassessment. AHPwas alsoused, incombinationwithUtilityTheory, byHsuehet al. (2007) todevelopamulti-criteriariskassessment model forconstruc-tionjoint-ventures.Themodeldidnotprovideaprojectriskassessment; it calculated the expected utility of the project inquestion. Hence, the higher the expected utility value, the lowerthe project risk level.In order to improve risk modelling, Aven et al. (2007)discussed the nature of risk and argued that some risks are moremanageable than others. This means that the chance of reducingthe downside effect of one risk may be larger than other risks.Toreflect thisidea, theconcept of riskmanageabilitywasintroduced. According to Aven et al. (2007), an alternative withmediumrisklevelandlowmanageabilitycouldeventuallyberiskierthananoptionwithhighriskandhighmanageability.Althoughmanageabilityisakeyissuetobeconsideredwhenassessingrisk, Avenet al. (2007) didnot suggest aspecificmethodology for assessing it. Similarly, Dikmen et al. (2007b)addressedtheissueofriskmanageabilityorcontrollability,but in a rather different manner. They argued that the capabilityofacompanyin managingproject risksshouldbeconsideredwhen assessing them. As a result, they consider the experienceof aconstructionfirmasaninfluencingfactor that mitigatesproject risklevel. Actually, Dikmenet al. (2007b)dealt with5 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights froma literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004manageabilityasanattributeof thecompanyrather thananintegratedattributeofriskitself.Thisisadifferentinterpreta-tionfromthemanageabilityconceptofAvenetal.(2007).Inanother proposal to improve modelling risk, Cagno et al. (2007)discussed risk controllability again but froma differentperspective. They adopted the PI risk model and defined riskcontrollability as the ratio between the expected risk impact,and monetary equivalent, before and after applying mitigatingactions. Inotherwords, riskcontrollabilityisdealt withasatool for justifying mitigating actions economically. The proposedmodel was aimed to be used at a company level; at a project levelrisk is dealt with as a variation in project parameters. The PI riskmodel was again subject to another improvement by Zeng et al.(2007)throughaddingathirddimension; Factor Index(FI).This dimensionreflects thesurroundingenvironment andtheinfluences betweenthe identifiedrisks. Inother words, it isrepresentingthe complexityof risk assessment withina realproject context. Riskismodelledas: R = L S FI whereLstandsforlikelihood ofoccurrence and Sstandsfor severityofrisk impact. The complexity of risk assessment was also addressedby Ackermann et al. (2007). According to them, the interactionproject risks can cause the most damage to a project. Hence, theysuggestedlookingat project risksasanetworkof interrelatedpossible events, which they referred to as risk systemicity. Theyarguedthat duetotheir interactions, project risks canformaportfoliowithanoverall impact greater thanthesumof theindividual riskimpacts. Besides, theyexplainedthat oneriskoccurrence may result in reinforcement of the likelihood of otherrisks occurring. Hence, they urged for considering this complexchainof outcomeswhenassessingrisks aswell as therisksthemselves. Indeed, such a holistic appreciation of the complexityof risk assessment is crucial for obtaining realistic results.Zeng et al. (2007) used FST to handle subjectivity inconstructionriskassessment anddeployedAHPtoprioritiserisks and derive relative weights for them. Similarly, Zhang andZou(2007)combinedthestrengthsofFSTandAHPwithin,what they call, a FuzzyAHP approach for assessing risks injoint venture construction projects in China. It is worth notingthat bothAHPandFSThavelimitations andutilisingthemtogetherdoesnotnecessarilyovercometheirlimitations.LikeZenget al. (2007), Zhang(2007) insistedthat assessingriskshould not neglect the surrounding environment. He argued thatproject environment has a mitigating effect on risk assessmentwhichislargelyneglectedwhenusingstatisticalmethods.Headvocatedconsideringthespecificnatureof aproject whenassessing risk and proposed the concept of project vulnerabil-ity. Project vulnerability has two distinct dimensions: theexposureofaproject toarisk; andthecapacityofaprojectsystemto copewithriskimpacts.It is a very important papertrying to push research towards a more realistic risk assessment.However, it did not suggest a method for assessing vulnerabil-ity or incorporating in risk assessment or risk model. Yet, it wasabasisforaninnovativeapproachforhandlingproject riskssuggested by Vidal and Marle (2012) as will be discussed later.Whileestimatingconstructionproject mark-upthroughCase-Based Reasoning, Dikmen et al. (2007a) considered risk aprojectattributewhichcanbehandledthroughcalculatinganaccurate contingency sum. In another paper Dikmen et al.(2007c) used the Analytic Network Process (ANP) for projectappraisal andalsoaddressedriskasamajorproject attribute.Zou et al. (2007) investigated the key risks affecting construc-tionindustryinChina. Risks wereidentifiedandprioritisedaccordingtotheirsignificancewhichistheimpactonprojectobjectives such as cost, time, quality, safety and environmentalsustainability. In this study, the PI risk model was adopted andproject risklevel wasdefinedastheaverageof theassessedrisks. Risk significance, however, was defined differently in Hanet al. (2008). Athreedimensional riskmodel, SignificanceProbabilityImpact, was proposed with risk significance anddefinedas the degree towhicha practical expert feels riskintuitively. This includes a general recognition of risk, thedifficulty of gaining information and implementing managementskills, the degree of indirect or potential loss and the relationshipbetweenproject profitabilityandtheanalyst'sattitudetowardrisk.AlthoughthepaperpresentsaDSSforassessingtherisklevel ofspecificriskpath, source-event, orproject scenario, itdoes not suggest any mechanismfor aggregating individualriskassessmentsor generatingproject risklevel. Zayedet al.(2008), though, proposedariskmodel that calculatesprojectrisk level for prioritising a set of projects. Project risk level wasdefined as the product of two risk indices: R1 that measures theriskson amacrolevelof theprojectandR2 that measures themicro level ones. Hence, project risk is modelled as: R =R1 R2. Both R1 and R2 are defined as weighted sums of riskeffectsassessedbyindividual experts, withAHPdeployedtogenerate importance weights of the risks. In this paper thePIrisk model was not adopted. Instead, a collective score, risk effectEi, is used. Obviously, the method of generating project risk levelneglects the interdependencies between risks.Despite the dominance of the analytical approaches thatutilise FSTand AHP, the stochastic methods continued toappear in the literature. Cioffi and Khamooshi (2009) presenteda statistical methodology for combining risk impacts andgenerating an overall impact, at a given confidence level, whichlead to an appropriate contingency budget. The paper adoptedthe PI risk model and defined risk impact as the incurred costof risk occurrence. This methodology had a weakness inrequiring the probabilities of occurrence to be averaged in orderto perform the aggregation at certain confidence levels. In fact,another stochastic method, the MCS-based methodology devel-opedbythe WashingtonState Department of Transportation(WSDOT) for estimating project cost, was presented by Molenaar(2005). This methodology adopts the PI risk model and measurerisk impact by the extra cost incurred because of a risk happening.Similartotheprevioustwopapers, Luuet al. (2009)didnotdeal withriskasaproject attribute;theymodelledriskastheprobabilityofconstructionproject delay. Moreover, theyusedBayesian Belief Network (BBN) to model the relationshipsbetweentherisksthat causeproject delayandtoquantifytheprobabilityofaconstructionproject delay. Funget al. (2010)developed an Excel-based tool to assess project risk level from asafety perspective. Risk was modelled as a multiplication of anaccident frequency and its severity. Severity, or risk impact, wasdefined as the sumof three risk impacts: extra time, extra cost and6 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights from a literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004personal injury. Theresearchers adoptedthemethodologyofunifying different types of risk impact proposed by Larsson andField(2002). Thethreetypesofimpact werenormalisedintoscores between 01. Hence, the maximum possible severity of arisk is 3. Although the tool was able to rank the risks according totheirscores, aggregatingriskscoresforgeneratingprojectrisklevel was not considered. Mojtahedi et al. (2010), in turn, triedto extend the conventional project risk assessment by includinghealthandsafetyandenvironmentalaspects.Theyappliedthemultiple attribute group decision making technique (GTOPSIS)for collating different opinions of risk experts to prioritise risks.Theyalsoconsideredriskimpact onprojectcost,durationandhealth, safety and environment. The complexity and subjectivityof construction risk assessment was handled by Nieto-Morote andRuz-Vila(2011) throughcombiningthestrengthof AHPandFST. Althoughsuchacombinationwassuggestedpreviously,Zeng et al. (2007) and Zhang and Zou (2007) for instance, theresearcherstriedtoimproverisk modellingandassessment viaadopting the risk model of Cervone (2006) which was discussedearlier inthepaper. Theyusedlinguistictermstoassessrisklikelihood, impact and discrimination. They also used the fuzzyarithmetic average to aggregate the assessments of differentexperts and the fuzzy multiplication rule to generate a project risklevel. The FST appeared again in Lazzerini and Mkrtchyan (2011)for tackling the subjectivity and complexity of risk assessment, duetotheinterdependenciesandcausalitiesbetweenrisks, throughusing the Extended Fuzzy Cognitive Maps (E-FCMs). They alsousedtheFuzzyCognitiveMaps for synthesisingtheopinionsofdifferent expertsandsupportinggroupdecisionmakingandriskassessment. Thecomplexityof riskmanagement wasalsodiscussed in Marle and Vidal (2011) who argued that the existingtechniqueshavealimitationinconsideringtheinteractionsbe-tweenriskswhenassessingthem. Theystatedthat theexistingtools generallydeal withinterdependent risks as if theywereindependent. Hence, the researchers dealt with project risks froma project complexity-based perspective and focused on capturingthe interactions between risks that affect the various elements oftheproject. Infact, project complexityhasreceivedenormousattention over the last two decades. Researchers have investigatedthe concept of project complexity and the link between complexityand risk. Very often, risk was considered as an element of projectcomplexity or as a result of it. Actually, the ongoing problem oflacking enough information is a major cause of the complexity ofprojectrisksassessment.Hashemietal. (2011)investigatedtheproblemoflackingenoughdatatousetheparametricstatisticaltools for estimating risk probability and impact. They suggestedusing a nonparametric technique: Bootstrap technique, to generateinterval values with smaller standard deviations. Fang and Marle(2012), however, tackled the complexity issue in a differentmanner. They represented the complex interactions between risksthroughnetworksandusedthedesignstructurematrix(DSM)for modellingthecasual relationshipsbetweentherisks. TheyalsousedtheAHPfor evaluatingriskinteractions. Simulationtechniqueisusedtoanalysetheriskpropagationphenomenainthenetworkandtore-evaluateriskprobabilitiesandimpactsindifferentscenarios.Moreover,thecomplexityoftheassessmenttask was handled through an integrated DSS that enables decisionmakers to conduct a comprehensive risk management, identifica-tion, analysis and control, effectively and efficiently.Finally, risk assessment was handled in a rather innovativemanner by Vidal and Marle (2012). They utilised the concept ofproject vulnerability of Zhang (2007) and the Systems Theoryandarguedthat analysingproject risks couldbebetter con-ductedthroughfocusingonthe existingweaknesses of theproject systems. Hence, instead of assessing risks directly, vul-nerabilitymanagement enables assessingthe weaknesses ofthesystemsresponsiblefor managingproject risks. Vulnera-bilitymanagementwasrecommendedasapromisingtoolforhandling the complexity of the projects and project risk analysis.Indeed this is a rather innovative way of looking at project risks;it may open doors for new ways of analysing project risks andevaluating project performance.Totheresearcher'sknowledge,theaboveliteraturereviewreflects the main development trends in project risk modellingand assessment over the last five decades. The next section ofthepaperprovidesathoroughanalysisanddiscussionofthefindings of this review.3. Analysis and discussion of risk modelling andassessment literature3.1. Risk modellingReviewingtheliteratureshows that constructionriskhastraditionally been perceived as the variance of cost or durationestimation. Gradually, there has been a shift in perceptiontowardsseeingitasaprojectattribute.Asaprojectattribute,riskismainlymodelledasamultiplicationof probabilityofoccurrenceandImpact. It isevident that thePI riskmodeldominatestheliterature. However, aconsiderablenumber ofimprovement proposals canbe appreciated. These improve-ment proposals can be summarised by:Predictability: Charette (1989) proposed adding predict-ability as athirddimensiontothePI riskmodel. Thisimprovement was braised by Williams (1996).Exposure: Jannadi and Almishari (2003) added the extent ofexposure to risk as a third dimension to the PI model;Discrimination: Cervone (2006) calledonconsideringtheinterdependencies between risks through reducing their inde-pendent scores generated by the PI model;Manageability: Aven et al. (2007) argued that some risks aremore manageable than others and urged analysts to considerthisfact whenassessingrisks. Incorporatingriskmanage-ability was also suggested by Dikmen et al. (2007b), but asan influencing factor that could mitigate the overall projectrisk level;Controllability: Cagno et al. (2007) considered riskcontrollability as a ratio between the expected risk impactsbefore and after applying specific mitigation actions;Factor Index: Zeng et al. (2007) addressed the influence of thesurrounding environment and the interdependencies betweenthe identified risks by incorporating the factor index (I) as athird dimension in the PI risk model;7 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights froma literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004Project Vulnerability: Zhang (2007) advocated extendingproject risk analysis process by incorporating project vulner-ability in order not to neglect the mediating effect of projectenvironmentonriskimpact.Project vulnerabilitywasinno-vatively deployed by Vidal and Marle (2012) who argued thatanalysingproject risks couldbebetter conductedthroughfocusing on the existing weaknesses. They introduced vulner-ability management which enables assessing the weaknessesof the systems responsible for managing project risks;Significance:Hanetal.(2008)addedrisksignificance asathirddimensiontothePImodel inordertoreflect theunique nature of risk and the intuition of risk analysts whenassessing a risk.The above development proposals have suggested differentapproaches. Avenet al. (2007) andZhang(2007) calledonconsideringriskmanageabilityandprojectvulnerabilitywhenassessingriskwithoutrecommendingaspecificmethodtodothat. Vidal andMarle(2012) suggestedfocusingonprojectweaknesses for managing its risks. Jannadi and Almishari(2003), Cervone(2006), Zenget al. (2007), andHanet al.(2008), however, extendedthe PI model andincorporatedadditional dimensions. Dikmen et al. (2007b), though, incorpo-rated risk manageability as an influencing factor on project risklevel. Whatever the format of the improvement was, the aim wasto reflect the complexity of risk assessment and to make the PImodel more reflective through incorporating the unique natureoftheriskandappreciatingtheinfluenceof thesurroundingenvironment and the interdependencies between risks.Among these different approaches, the notion of extending thePI risk model by incorporating additional explicit parameter(s)seems practical and convenient. For instance, the proposedincorporation of risk manageability by Dikmen et al. (2007b)as an influencing factor on project risk level might not be veryhelpful. It wouldbedifficult for adecisionmaker toprovideanaccuratefigurethat reflects his or her experience, or thecompany's experience, in controlling project risks collectively; itmay be much easier for the analyst to assess his or her experiencein managing individual risks. Thus, when aggregating individualrisk assessments, the manageability of project risks will beeffectivelyestablishedandincorporated. Theauthoradvocatesthe idea of extending the PI risk model to incorporate additionalparameters capableof reflecting thetrue nature of therisk andtheexperienceof themanagement teaminhandlingrisk, thecomplexity of project systems that may affect risk assessment andtheinteractionbetweenprojectrisks. Suchan extensionwouldprovidethebasisfor adetailedandrealisticriskassessment.Moreover, such a model, that incorporates the influencing factorson risk assessment, comes into alignment with the complexity-drivenapproachesincreasinglyusedforassessingproject risk.Hence, the additional parameters should clearly reflect:Theuniquenatureofariskandtheexperienceoftheriskmanagement team in controlling its impact and mitigating it;The interdependencies between the identified risks; andTheeffect ofthesurroundingproject environment onriskprobability and impact.Theadditional parameterswill functionasmitigationcoef-ficients withproportional values between0and1. Theyareresponsible for reducing the maximumrisk assessment, generatedbythePImodel,aftertakingtheabovepointsintoconsider-ations. Hence, the parameters require very clear definitions asso-ciated with specific value ranges.3.2. Construction risk assessmentWhen examining the published literature of project riskassessment, two levels of analysis can be recognised: risk assess-ment and project risk level estimation.3.2.1. Risk assessmentDifferent approaches have been adopted for assessingproject risks. At first, researchers used statistical methodsbasedonPTfordealingwithdurationriskorcost risk. Thisapproach perceived risk as an estimation variance. In alignmentwiththisperception, objectiveprobability, or frequency, hasbeenalways sought after. Gradually, manyresearchers con-cludedthat humanfactors, intuition, professional experienceand personal judgement were essential for risk assessment.Toreflect this conclusion, FSTwas introducedas a viablealternative for handling subjectivity in construction risk assess-ment. Besides the subjectivity issue, researchers were faced bytheever increasingcomplexityinriskassessment duetothegrowingcomplexityinconstructionprojects (TahandCarr,2000).Todealwiththischallenge,AHPwasperceivedasaneffectivetoolforhandlingthecomplexityinconstructionriskassessment. It has provided a systematic approach to structuringriskassessmentproblemsbyprovidingalogicalapproachforassessing risk impacts and allocating importance weighting. Infact,thecomplexitysurroundingriskassessment hasattractedhugeattention. Researchershavetriedvariousapproachesforrepresentingtheinterdependenciesbetweenproject risksandreflecting the complexity of the surrounding environment(Ackermann et al., 2007;Baloi and Price,2003;Cagno et al.,2007; Choi et al., 2004; Diekmann, 1992; Hanet al., 2008;Hastak and Shaked, 2000; Kangari and Riggs, 1989; Lazzeriniand Mkrtchyan, 2011; Mustafa and Al-Bahar, 1991;Nieto-Morote and Ruz-Vila,2011;Thomaset al., 2006;Zenget al., 2007; ZhangandZou, 2007). Yet, AHPhas enjoyedpopularityrecently. ThewideadoptionofAHPoverthelasttwodecadesreflectedthechangeinriskperceptionfromanestimation variance into a project attribute. In fact, Laryea andHughes (2008)referred to a paradigm shift in risk assessmentfromclassicalism; using PT-based and simulation tools,towards conceptualism; using analytical tools. It is clear thatthe change in risk perception was coupled with a change in thetools for assessing risks. According to Laryea and Hughes(2008), however, the paradigm shift did not result in a greateradoption of the analytical tools by professionals.They arguedthat the take-up of the available risk assessment tools bypractitioners is quite limited. The limited adoption of riskassessment tools does not hamper their potential. Researchfindings tell us that people who use advanced risk assessmentmethods appreciate their potential and feel positive about their8 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights from a literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004benefits (Simister, 1994). Indeed, the existinggap betweentheoryandpracticeshouldgalvanisetheefforts towards in-vestigating more suitable theories and approaches that appeal topractitionersandreflect their experienceandpractice. More-over, simplicityisakeyfactorforencouragingprofessionalstouseriskassessment toolsaswill bediscussedlaterinthepaper. Furthermore, riskassessment must befacilitatedbyauser friendly DSS that structures the complexity of the task andallows professionals to implement their strategies and tactics ina rather easy and transparent manner.In terms of risk impact assessment, it is striking howneglected the analysis of project quality risk is; the attention isstill focused on cost risk or duration cost. In fact, literature lackssufficient research on assessing risk impact on project quality orother strategic objectives. Subsequently, it lacks an assessmentmethodologythat iscapableofunderstandingriskimpact onall project successobjectives. Thereasonforsuchscarcityisattributedtothe lackof a commonscale (Williams, 1995).Yet, the most convenient common scale is thought to be the riskcost (Franke, 1987; Williams, 1995). Actually, riskcost hasbeenusedas a riskimpact measurement scale bydifferentresearchers (Ben-DavidandRaz, 2001; Cagnoet al., 2007;Cioffi and Khamooshi, 2009; Fan and Yu, 2004; Franke, 1987;Molenaar, 2005). Nevertheless, none of them used risk cost forassessingriskimpact ondifferent project objectives. Inthecasesusingmonetaryequivalentsforquantifyingriskimpact,theassessment outcomeis anexpectedcontingencysumtocover theriskor anexpectedcost of mitigatingits impact.Hence, it is worthinvestigatinganassessment methodologythat enables assessing risk impact of specific project objectivesusing risk cost as a common scale. This could successfully leadto a comprehensive riskassessment and arealisticassessmentof project risk level. Although one may question the suitabilityof using risk cost as a scale for assessing intangible element likequalityriskfor instance, the authorbelievesthat itcouldbe apractical approach. Practitionerswhopriceresidual risks, theunknown unknowns, and provide a contingency allowancefor coveringthemshouldbeabletopricetheimpact of theknownunknowns, therisks, onproject quality. Quantifyingrisk, pricing it, might be more appealing for practitioners due tothefact that thevast majorityoftheexistingriskassessmenttools provide a risk rating, numerical scores or linguisticvariables, rather than an actual quantification. Truly, this mighttosomeextentexplainthelowtake-upofthesetools.Hence,using risk cost as a measurement scale may contribute to closingthe gap between theory and practice.3.2.2. Project risk levelDifferent approacheswereutilisedtoobtainaproject risklevel. Conventionally, researchersusedPT-basedtechniques,PERTandMCSmainly, for combiningthe probabilitydis-tributions of the activities' durations or costs and assessing theriskof project cost overrunor project delay. As mentionedearlier, theperceptionof project riskhas graduallychangedfrom an estimation variance into a project attribute. At first, riskwasdealtwithasaprojectattributethatcanbeanalysedataprojectlevel.Suchanapproachissufficientforanalysingtheriskof small projects, not large complexones (Deyet al.,1994). Increasingly, systematic and more sophisticated ap-proaches have been adopted to handle risks in the increasinglycomplexprojects. Project risks are systematicallyidentified,categorised and structured in different manners such as influencediagrams, Bayesian networks, fault trees and the hierarchical riskbreakdown structure which is believed to be the most commonlyused one. The sophistication of the risk structures reflected thecomplexityof the project systems andthe interdependenciesbetweentherisks.Aggregatingtheindividualriskassessmentsacrosstheirstructuresforgeneratinganestimateoftheprojectrisk level was a challenge. In most of the cases the aggregationmechanismhasaveragedtheindividual riskassessments. Thefuzzy averagingrule for instance has beenusedwidelyforaggregatingtheindividualriskassessmentsinlinguisticterms.Averaging might not be the best option for obtaining a realisticriskassessment.Inothercases,projectrisklevelwasobtainedas theweightedsumof theindividual assessments. Suchanapproach has a limitation of assuming that the aggregated risksare independent. Utility Theory was also employed for estimatingproject risk level. Project utility represented the attractiveness orthe risk level of a project; the smaller the project utility the biggerthe risk level. In this case, the overall project utility was derivedeither byasimpleor aweightedsumof individual utilities.Again, this approach has a limitation of being over-simplistic dueto the assumption of independence between risk factors (Dikmenet al., 2004). One could argue that the key for obtaining a realisticprojectrisklevelisstartingwithrealisticriskassessmentsandthen deploying an effective aggregation mechanismthat keeps theanatomyof theindividual assessments. Therefore, researchinganovel approachfor assessingrisksandutilisinganeffectiveaggregation rule is crucial for a successful project risk assessment.Besides, producing novel alternatives should be appealing for theend users. Hence, understanding the reasons of the low take-up ofthe existing tools is essential for devising better alternatives andclosing the existing gap between the theory and practice of riskassessment.4. Surveying the actual practice of risk assessmentWhile project management literature is rich in papersaddressingriskmanagement, fewpapers haveresearchedtheactual practiceof riskassessment andinvestigatedthepracti-tioners' points of view regarding the available tools and DSSs.Nonetheless, theexistingliteraturegivesagoodinsight abouttheseissues. Tahet al. (1994)researchedthepracticeofcostestimation and the usage of statistical tools for assessing the riskof cost overburdeninconstructionprojects. Theyfoundthatall contractors did not perform any form of statistical analysis ofrisk. Contractorsexplainedsuchanattitudebyindicatingthatstatistical tools could not effectively represent the unique natureof construction risk. Akintoye and MacLeod (1997) investigatedthe actual practice of RM in the UK construction industry. Theyarguedthat experience andintuitionwere keytools for riskassessment. Moreover, they found that sophisticated quantitativetools were not widely used for assessing risk. Similar results wereobtainedbyShen(1997) whoadministeredasurveyof risk9 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights froma literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004managementpracticeinHongKong.Thesurveyrevealedthatexperience and subjective judgement were the most effective andwidely used tools for managing risks. The results also showed averylimitedusage of quantitative techniques due tolimitedunderstanding and lack of experience in such methods. Based onthese studies and other similar ones like Potts and Weston (1996)andJacksonetal. (1997), WoodandEllis(2003)investigatedthe actual practice of RM in the UK construction industry. Wherethe previous studies administered questionnaires to collect data,WoodandEllis (2003) conductedin-depthinterviews inanattempt toget deeperinsightsandgeneratequalitativeresults.Their research findings, to a large extent, confirmed many of thefindings of the previous questionnaire-based research.Baker et al. (1998) surveyed the most successful qualitativeand quantitative risk analysis tools inconstruction and OilandGasindustries. Theresearchersconcludedthat themostfrequently used qualitative tools, and the most successful ones,arepersonalandcorporateexperienceandengineeringjudge-ment. Regardingquantitativeriskanalysis, it wasfoundthatthe Expected Monetary Value (EMV), break-even analysis,scenario analysis and sensitivity analysis were the most widelyused. According to them, these tools are mainly used for handlingfinancial risks. Moreover, theysuggestedthat thismayreflectthepreoccupationofconstruction industrywithfinance. It isnotable that these tools are not sophisticated in general. This maysuggest that practitionerstendtousesimplequantitativetoolsforsupportingtheirexperienceandjudgementwhenassessingconstructionrisks. Infact, this suggestionwas confirmedbyWoodandEllis(2003). AccordingtoWoodandEllis(2003),reliance on personal judgement and experience is prevailing; RMis usually conducted using simple tools like checklists and riskregisters; usingcomplexanalysistechniquesislimitedduetoscepticism about their usefulness; cost is used as a measure forrisk impact; and RM is believed to benefit particularly in projectbudgeting and contingency estimation. For comparison and con-firmation reasons, Lyons and Skitmore (2004) conducted asurveyamong200constructioncontractors inQueenslandinAustralia. In general, the results were consistent with the resultsof the previous studies and the results of another three surveysindifferent timesandcountries(Bakeret al., 1999a; RazandMichael, 2001; Uher and Toakley, 1999). The results confirmedthat intuition, judgement and experience are the most frequentlyusedriskassessment techniques. Moreover, theyshowedthatqualitativemethods for riskassessment wereusedmorefre-quentlythanquantitativeor semi-qualitativemethods. Similarresults were presented by Dikmen et al. (2004) who found thatpersonal judgement and experience were the key tools forassessing risk qualitatively. They also found that sensitivityanalysis and decision tree were the most commonly used tools forquantitative assessment. In another study, Warszawski and Sacks(2004) argued that sensitivity analysis was the most commonlyusedtool for riskassessment inconstruction. Theyexplainedtheir results by arguing that the more sophisticated methods werenot widely used because they required detailed input informationwhich might not be available in every case.The common theme emerging from the above studies showsthat the actual practice of risk assessment is, very much, basedonexperienceandpersonal judgement. Besides, practitionersdo not use the available tools or DSSs extensively; they prefertoconduct riskanalysis inarather simpleandpersonalisedmanner. Moreover, it seems that theypreferto use the simplequantitative tools to perform a quantitative risk analysis. Such aconclusion might need to be always considered when introducingany alternative to the existing tools. Laryea and Hughes (2008)concluded that the problem of the limited adoption of the existingtools can be dealt with through devising risk assessment method-ologies that appreciate the actual practice in construction industryand reflect what practitioners do in reality. In fact, they suggestedthat introducing new models or methods may not necessarily beuseful. Although such a position can be appreciated, it should notprevent researchers frominvestigatingnovel alternatives anddevising usable DSSs that simulate the actual practice of the endusersandallowthemtoexpresstheirpersonaljudgementandutilisetheir cumulativeexperience. Yet, thetools needtobereliable and based on a firm theoretical base.5. Summary and conclusionsThis paper presentedareviewof theriskmodellingandassessment literaturepublishedover thelast fivedecades. Itfocused on the development of risk modelling and identified anumber of proposals for improvingthe prevailingPI riskmodel. It alsodiscussedthe different contributions towardsinvestigatingvarioustheoriesandtechniquesfor riskassess-ment. Thepresentedreviewrevealssignificant results, whichmay be summarised as follows.It was found that the PI risk model still prevails, yet effortshave increasedrecentlyfor improvement. The improvementproposalshaveacceleratedrecentlyandcomeintoalignmentwiththe increasingcomplexityof constructionprojects andconstructionrisks. Moresophisticationisbeingincorporatedintoriskmodellingandassessment inanattempt toreflectdifferentaspectsofcomplexityliketheinterdependenciesbe-tween risks and the interaction between themand the surroundingproject environment. The improvement proposals, however, arenot comprehensive enough to capture, at the same time, the riskcharacteristics, the interdependencies between risks, the interac-tion with the complex project environment and the experience ofthemanagementteams.Indeed,understandingprojectvulnera-bility is crucial for advancing construction risk assessment. Thisisanovel approachforemployingthepractical experienceofconstructionprofessionals for aidingriskassessment. It wasfound that there was an evident shift from perceiving risk as anestimation variance towards considering it as a project attribute.Thisshiftwasaccompaniedwith ashiftin assessingrisk fromusing PT-based approaches towards using more analytical ones.Infact, theanalytical approaches, that employFSTandAHPmainly, haverecentlydominatedtheliterature. Progressively,sophisticated DSSs are being devised to facilitate the analyticalandstatistical toolsandtohandlethegrowingcomplexityofrisk assessment. These DSSs enable conducting comprehensiveanalysesincludingriskidentification, riskassessment, assess-ment aggregationandproject risk level estimation.Thereviewhas confirmed what was previously mentioned that the literature10 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights from a literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004lacksacomprehensiveriskassessmentframeworkwhichcon-siders the different types of impact of a risk on different projectobjectives simultaneously. Suchaframeworkis essential forobtaining realistic risk assessments which is the first step towardsgenerating a realistic project risk level. Lacking a common scalewas blamed for this limitationin literature. Althoughmanyauthorshaverecommendedriskcostasacommonscaleforassessingriskimpactonspecificproject objectives(ChanandAu, 2008; Dey, 2001; Franke, 1987; Paek et al., 1993; Sanchez,2005; Williams, 1993, 1995), it was found that risk cost has neverbeen used systematically forsuch apurpose. Thereview dem-onstrated that we lack an effective aggregation rule that generatesa realistic project risk level without compromising the anatomyof theindividual riskassessments. Theconventional rules ofaggregation, the averaging or the weighted sum, are not alwayssuitable ways for obtaining a representative project risk level dueto their underlying assumptions.The paper has also reviewed the actual practice of riskanalysisaspublishedintheliterature. Thefindingsrefertoaheavy reliance on practical experience and professional judge-ment when assessing construction risk. Moreover, the adoptionof the available tools and DSSs is quite limited. Truly, this reviewhasdemonstratedaremarkablecontributionof theresearcherstowardsadvancingriskmodellingandassessment. It isunfor-tunate that there is still a wide gap between theory and practice.Yet, theexistingbodyofknowledgedemonstratesafirmbasisfrom which toinvestigatenovelalternatives that can bridgetheexisting gap between theory and practice. This paper recommendsextending the PI risk model and incorporating additionalparameters toreflect thenatureof therisk, theexperienceofriskanalysts, theinterdependencies betweenproject risks andtheinfluenceofproject environmentonrisk assessment.It alsosuggestsusing riskcost,as acommon scalefor measuring riskimpact onvarious project objectives, withinananalytical ap-proach which structures and facilitates the experience and personaljudgement of construction professionals for assessing constructionrisk. Risk cost, a percentage of project initial cost for instance, isbelieved to be a convenient and practical measure of risk impactas it is a common language understood by all construction parties.The simplicity of using risk cost for assessing risk and the facil-itation of the practical experience are believed to be key virtues ofthe suggested approach. This may appeal to construction profes-sionals andparticipate inadvancingthe real practice of riskassessment. This research is part of a wider research project aimedat rethinkingconstructionriskmodellingandassessment andfacilitating closing the existing gap between theory and practice ofconstruction risk assessment.AcknowledgementThe authoris hugelygratefulfor the constructive feedbackandsuggestionsprovidedbythethreeanonymousreviewersandtheeditorial boardoftheInternational JournalofProjectManagement. Theyhavebeenextremelyuseful inimprovingthe paper to its current version.Appendix A. Key papers in project risk modelling and assessment literature reviewPhases Number of papers Year Author(s) Journal Main tools & theoriesThe 1980s 8 1983 Barnes IJPM PT1983 Chapman and Cooper JORS PERT, OR tools1983 Diekmann JCEM MCS1985 Cooper et al. IJPM PT1986 Beeston CME MCS1987 Clark and Chapman EJOR PT1987 Franke IJPM LR, Cognitive argument1989 Kangari and Riggs IEEETEM FSTThe 1990s 18 1990 Hull IJPM MCS, PERT1990 Al-Bahar and Crandall JCEM Influence diagramming, MCS1990 Yeo JME PERT1991 Mustafa and Al-Bahar IEEETEM AHP1992 Diekmann IJPM MCS, FST, Influence diagramming1992 Huseby and Skogen IJPM Influence diagram, MCS1993 Paek et al. JCEM FST1993 Tah et al CSE FST1994 Dey et al. IJPM AHP, PT1994 Riggs et al. COR AHP, Utility Theory1995 Williams EJOR Literature review (LR)1995 Zhi IJPM AHP1996 Williams IJPM LR, Cognitive argument1996 Wirba et al. ECAM FST1998 Dawood CME MCS, PERT1998 Tavares et al. EJOR PT1999 Mulholland and Christian JCEM PT, PERT1999 Ward IJPM LR, Cognitive argument11 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights froma literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004(continued)Phases Number of papers Year Author(s) Journal Main tools & theoriesThe new millennium 13 2000 Chapman and Ward IJPM LR, Cognitive argument2000 Hastak and Shaked JME AHP2000 Tah and Carr CME FST2001 Baccarini and Archer IJPM LR2001 Ben-David and Raz JORS Optimisation through Microsoft Excel VBA2001 Dey MD AHP, Decision tree2001 Xu and Tiong CME Stochastic programming2001 Tah and Carr JCCE FST2002 Patterson and Neailey IJPM LR, risk register2003 Baloi and Price IJPM LR, Cognitive argument2003 Jannadi and Almishari JCEM LR, Cognitive argument2003 Ward and Chapman IJPM LR2004 Choi et al. JCEM FSTPost 2005 29 2005 Molenaar JCEM MCS2005 Shang et al. ECAM FST2006 Cervone OCLC S&S LR, Cognitive argument2006 Dikmen and Birgonul CJCE AHP2006 Poh and Tah CME Influence networks2006 Thomas et al. CME FST, Fault tree2007 Ackermann et al. JORS Theoretical argument, Risk Register2007 Aven et al. RESS LR, Cognitive argument2007 Cagno et al. RM LR, Cognitive argument2007 Dikmen et al. a AIC Case-Based Reasoning, Utility Theory2007 Dikmen et al. b IJPM FST, Influence diagramming2007 Dikmen et al. c CJCE ANP2007 Hsueh et al. AIC AHP, Utility Theory2007 Zeng et al. IJPM AHP, FST2007 Zhang and Zou JCEM AHP, FST2007 Zhang IJPM LR, Cognitive argument2007 Zou et al IJPM LR, Cognitive argument2008 Han et al. AIC LR, AHP2008 Zayed et al. IJPM AHP, Questionnaire2009 Cioffi and Khamooshi JORS PT2009 Luu et al. IJPM Bayesian Belief Networks (BBN)2010 Fung et al. IJPM Microsoft Excel2010 Mojtahedi et al. SS GTOPSIS2011 Hashemi et al. JCEM Bootstrap2011 Nieto-Morote and Ruz-Vila IJPM FST, AHP2011 Lazzerini and Mkrtchyan IEEESJ Fuzzy Cognitive Maps2011 Marle and Vidal RED Cognitive argument, Clustering heuristics2012 Fang and Marle DSS AHP, Simulation2012 Vidal and Marle Kybernetes Cognitive argument, Systems TheoryAbbreviation JournalAIC Automation in ConstructionBAE Building and EnvironmentCJCE Canadian Journal of Civil EngineeringCME Construction Management and EconomicsCSE Computing Systems in EngineeringCOR Computers and Operations ResearchDSS Decision Support SystemsECAM Engineering, Construction and Architectural ManagementEJOR European Journal of Operational ResearchIEEETEM IEEE Transaction on Engineering ManagementIEEESJ IEEE Systems JournalIJPM International Journal of Project ManagementJCEM Journal of Construction Engineering and ManagementJCCE Journal of Computing in Civil EngineeringJME Journal of Management in EngineeringJORS Journal of the Operational Research SocietyJSS The Journal of Systems and SoftwareMD Management DecisionOCLC S&S OCLC Systems & ServicesAppendix (continued)12 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights from a literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004(continued)Abbreviation JournalRED Research in Engineering DesignRESS Reliability Engineering and System SafetyRM Risk ManagementSS Safety ScienceReferencesAckermann, F., Eden, C., Williams, T., Howick, S., 2007. Systemic risk assessment:a case study. Journal of the Operational Research Society 58, 3951.Akintoye, A.S., MacLeod, M.J., 1997. Risk analysis and management inconstruction. International Journal of Project Management 15, 3138.Al Bahar, J.F., Crandall, K.C., 1990. Systematic risk management approach forconstruction projects. Journal of Construction Engineering and Management116, 533.Aven, T., Vinnem, J., Wiencke, H., 2007. Adecision framework for riskmanagement, with application to the offshore oil and gas industry. ReliabilityEngineering and System Safety 92, 433448.Baccarini,D.,Archer,R.,2001.Theriskrankingofprojects:amethodology.International Journal of Project Management 19, 139145.Baker, S., Ponniah, D., Smith, S., 1998. Techniques for the analysis of risks inmajor projects. Journal of the Operational Research Society 49, 567572.Baker,S., Ponniah,D.,Smith,S.,1999a.Riskresponsetechniquesemployedcurrently for major projects. Construction Management and Economics 17,205213.Baker, S., Ponniah, D., Smith, S., 1999b. Survey of risk management in majorUK companies. Journal of Professional Issues in Engineering Education andPractice 125, 94102.Baloi, D., Price, A.D.F., 2003. Modelling global risk factors affecting constructioncost performance. International Journal of Project Management 21, 261269.Barnes, M., 1983. How to allocate risks in construction contracts. InternationalJournal of Project Management 1, 2428.Beeston, D., 1986. Combiningrisksinestimating. ConstructionManagementand Economics 4, 7579.Ben-David, I., Raz, T., 2001. Anintegratedapproachfor riskresponsedevel-opment inproject planning. Journal of the Operational ResearchSociety1425.BSI-6079-4, 2006. Project management Part 4: guide to project managementin the construction industry. Institute, B.S., London.Cagno, E., Caron, F., Mancini, M., 2007. A multi-dimensional analysis of majorrisks in complex projects. Risk Management 9, 118.Carr, R.I., 1977. Paying the price for construction risk. Journal of the ConstructionDivision 103, 153161.Cervone, H.F., 2006. Project risk management. OCLC Systems and Services 22,256262.Chan, E.H.W., Au, M.C.Y., 2008. Relationship between organizational sizes andcontractors' riskpricingbehaviorsfor weather riskunder different projectvalues and durations. Journal of Construction Engineering and Management134, 673.Chapman,C., Cooper,D.F.,1983.Riskengineering:basiccontrolledintervaland memory models. Journal of the Operational Research Society 5160.Chapman, C., Ward, S., 2000. Estimationandevaluationof uncertainty: aminimalist first pass approach. International Journal of Project Management18, 369383.Charette, R.N., 1989. Software engineering risk analysis and management.McGraw-Hill, New York.Choi, H.H., Cho, H.N., Seo, J., 2004. Risk assessment methodology forunderground construction projects. Journal of Construction Engineering andManagement 130, 258.Cioffi, D.F., Khamooshi, H., 2009. A practical method of determining project riskcontingency budgets. Journal of the Operational Research Society 60, 565571.Clark, P., Chapman, C., 1987. The development of computer software for riskanalysis: a decision support systemdevelopment case study. EuropeanJournal of Operational Research 29, 252261.Cooper, D., MacDonald, D., Chapman, C., 1985. Risk analysis of aconstructioncost estimate. International Journal ofProject Management3,141149.Dawood, N., 1998. Estimating project and activity duration: a risk managementapproach using network analysis. Construction Management and Economics16, 4148.Dey,P.K.,2001.Decisionsupportsystemforriskmanagement:acasestudy.Management Decision 39, 634649.Dey, P., Tabucanon, M.T., Ogunlana, S.O., 1994. Planning for project controlthrough risk analysis: a petroleum pipeline-laying project. International Journalof Project Management 12, 2333.Diekmann, J.E., 1983. Probabilisticestimating:mathematicsandapplications.Journal of Construction Engineering and Management 109, 297308.Diekmann, E., 1992. Riskanalysis:lessonsfromartificialintelligence. Interna-tional Journal of Project Management 10, 7580.Dikmen, I.,Birgonul,M.T.,2006. Ananalytichierarchyprocessbasedmodelforriskandopportunityassessment ofinternational constructionprojects.Canadian Journal of Civil Engineering 33, 5868.Dikmen, I., Birgonul, M.T., Arikan, A.E., 2004. Acritical reviewof riskmanagement support tools. In: Khosrowshahi, F. (Ed.), 20th Annual ARCOMConference. Heriot-Watt University, Edinburgh, UK, pp. 11451154.Dikmen, I., Birgonul, M.T., Gur, A.K., 2007a. A case-based decision support toolfor bid mark-up estimation of international construction projects. Automationin Construction 17, 3044.Dikmen, I., Birgonul, M.T., Han, S., 2007b. Using fuzzy risk assessment to ratecost overrun risk in international construction projects. International Journalof Project Management 25, 494505.Dikmen, I., Birgonul, M.T., Ozorhon, B., 2007c. Project appraisal and selectionusing the analytic network process. Canadian Journal of Civil Engineering34, 786792.Edwards, P., Bowen, P., 1998. Riskandriskmanagement inconstruction: areviewandfuture directions for research. EngineeringConstructionandArchitectural Management 5, 339349.Fan, C.F., Yu, Y.C., 2004. BBN-basedsoftware project risk management.Journal of Systems and Software 73, 193203.Fang, C., Marle, F., 2012. Asimulation-based risk network model fordecisionsupport inproject riskmanagement. DecisionSupport Systems52,635644.Flanagan, R., Norman, G., 1993. Riskmanagement andconstruction. Wiley-Blackwell.Franke, A., 1987. Risk analysis in project management. International Journal ofProject Management 5, 2934.Friedman, L., 1956. Acompetitive-biddingstrategy. OperationsResearch4,104112.Fung, I.W.H., Tam, V.W.Y., Lo, T.Y., Lu, L.L.H., 2010. DevelopingaRiskAssessmentModelforconstructionsafety.InternationalJournalofProjectManagement 28, 593600.Gates, M., 1960. Statistical and economic analysis of a bidding trend. Journal ofthe Construction Division, ASCE 75107.Gates, M., 1967. Bidding strategies and probabilities. Journal of the ConstructionDivision, ASCE 93, 74107.Gates, M., 1971. Bidding contingencies and probabilities. Journal of the ConstructionDivision, ASCE 97, 277303.Gates, M., Scarpa, A., 1974. Rewardriskratio. Journal of theConstructionDivision, ASCE 100, 521532.Han, S.H., Kim, D.Y., Kim, H., Jang, W.S., 2008. Aweb-basedintegratedsystem for international project risk management. Automation in Construc-tion 17, 342356.Appendix (continued)13 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights froma literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004Hashemi, H., Mousavi, S.M., Mojtahedi, S.M.H., 2011. Bootstrap tech-nique for risk analysis with interval numbers in bridge constructionprojects. Journal of Construction Engineering and Management 137,600608.Hastak, M., Shaked, A., 2000. ICRAM-1: model for international constructionrisk assessment. Journal of Management in Engineering 16, 5969.Hertz, D.B., 1964. Risk analysis in capital investment. Harvard Business Review42, 95106.Hsueh, S.L., Perng, Y.H., Yan, M.R., Lee, J.R., 2007. On-line multi-criterionrisk assessment model for construction joint ventures in China. Automationin Construction 16, 607619.Hull, J., 1990. Application of risk analysis techniques in proposal assessment.International Journal of Project Management 8, 152157.Huseby, A., Skogen, S., 1992. Dynamicriskanalysis: theDynRiskconcept.International Journal of Project Management 10, 160164.Jackson, S.H., Griffith, A., Stephenson, P., Smith, J., 1997. Risk managementtools andtechniquesusedwhenestimatinginitial budgets for buildingprojects, 13thAnnual ARCOMConference. King'scollege, Cambridge123132.Jannadi, O.A., Almishari, S., 2003. Risk assessment in construction. Journal ofConstruction Engineering and Management 129, 492.Kangari, R., Riggs, L.S., 1989. Constructionriskassessment bylinguistics.IEEE Transactions on Engineering Management 36, 126131.Kendrick, T., 2003. Identifying and managing project risk. American ManagementAssociation, New York.Larsson, T.J., Field, B., 2002. The distribution of occupational injury risks inthe Victorian construction industry. Safety Science 40, 439456.Laryea, S., 2008. Risk pricing practices in finance, insurance, and construction,COBRA2008. TheconstructionandbuildingresearchconferenceoftheRoyal Institution of Chartered Surveyors, Dublin..Laryea, S.,Hughes, W., 2008. How contractors price riskin bids:theory andpractice. Construction Management and Economics 26, 911924.Latham, M., 1994. Constructing the team. HMSO London.Lazzerini, B., Mkrtchyan, L., 2011. Analyzing risk impact factors using extendedfuzzy cognitive maps. IEEE Systems Journal 5, 288297.Liu, J., Yang, J., Wang, J., Sii, H., 2002. Reviewofuncertaintyreasoningapproachesasguidancefor maritimeandoffshoresafety-basedassess-ment. Journal of United KingdomSafety and Reliability Society 23,6380.Luu, V.T., Kim, S.Y., Tuan, N.V., Ogunlana, S.O., 2009. Quantifying scheduleriskinconstructionprojectsusingBayesianbeliefnetworks.InternationalJournal of Project Management 27, 3950.Lyons, T., Skitmore, M., 2004. Project riskmanagement intheQueenslandengineering construction industry: a survey. International Journal of ProjectManagement 22, 5161.Marle, F., Vidal, L.A., 2011. Project riskmanagement processes: improvingcoordination using a clustering approach. Research in Engineering Design22, 189206.Merna, T., Al-Thani, F.F., 2008. Corporateriskmanagement. JohnWiley&Sons Inc.Mojtahedi, S.M.H., Mousavi, S.M., Makui, A., 2010. Project risk identificationand assessment simultaneously using multi-attribute group decision makingtechnique. Safety Science 48, 499507.Molenaar, K.R., 2005. Programmatic cost risk analysis for highwaymegaprojects. Journal of ConstructionEngineeringandManagement131, 343.Morin, T.L., Clough, R.H., 1969. OPBID: competitive bidding strategy model.Journal of the Construction Division 95, 85106.Mulholland, B., Christian, J., 1999. Risk assessment in construction schedules.Journal of Construction Engineering and Management 125, 8.Mustafa, M.A., Al-Bahar, J.F., 1991. Project risk assessment using the analytichierarchyprocess. IEEETransactionsonEngineeringManagement 38,4652.Nieto-Morote, A., Ruz-Vila, F., 2011. Afuzzyapproachtoconstructionproject risk assessment. International Journal of Project Management 29,220231.Paek, J., Lee, Y., Ock, J., 1993. Pricingconstructionrisks: fuzzysettheory.Journal of Construction Engineering and Management 119, 743756.Patterson,F.D.,Neailey,K.,2002.Ariskregisterdatabasesystemtoaidthemanagementofprojectrisk. InternationalJournalofProjectManagement20, 365374.Poh, Y., Tah, J., 2006. Integrated durationcost influence network for modellingrisk impacts on construction tasks. Construction Management & Economics24, 861868.Potts, K., Weston, S., 1996. Risk analysis, estimation and management on majorconstructionworks. OBInternational SymposiumNorthmeets SouthCommission W92, Durban. 522533.Raz, T., Michael, E., 2001. Use and benefits of tools for project riskmanagement. International Journal of Project Management 19, 917.Riggs, J.L., Brown, S.B., Trueblood, R.P., 1994. Integration of technical, cost, andschedulerisksinprojectmanagement.Computers andOperationsResearch21, 521533.Saaty, T.L., 1980. Analytic hierarchy process. Encyclopedia of Biostatistics..Sanchez, P.M., 2005. Neural-RiskAssessment Systemfor ConstructionProjects, Broadening Perspectives, ASCE 183. ASCE, San Diego,California.Shang, H., Anumba, C.J., Bouchlaghem, D.M., Miles, J.C., Cen, M., Taylor,M., 2005. An intelligent risk assessment system for distributed construc-tion teams. Engineering Construction and ArchitecturalManagement 12,391409.Shen, L., 1997. Project risk management in Hong Kong. International Journalof Project Management 15, 101105.Simister,S.J.,1994.Usageandbenefitsofprojectriskanalysisandmanage-ment. International Journal of Project Management 12, 58.Spooner, J.E., 1974. Probabilistic estimating. Journal of the Construction Division100, 6577.Tah, J., Carr, V., 2000. Aproposal for constructionproject riskassess-mentusingfuzzylogic.ConstructionManagementandEconomics18,491500.Tah, J., Carr, V., 2001. Knowledge-based approach to construction project riskmanagement. Journal of Computing in Civil Engineering 15, 170.Tah, J., Thorpe, A., McCaffer,R., 1993. Contractor project risks contingencyallocation using linguistic approximation. Computing Systems in Engineering4, 281293.Tah, J., Thorpe, A., McCaffer, R., 1994. A survey of indirect cost estimating inpractice. Construction Management and Economics 12, 3136.Thomas, A., Kalidindi, S., Ganesh, L., 2006. Modelling and assessment of criticalrisksinBOTroadprojects. ConstructionManagement andEconomics24,407424.Uher, T.E., Toakley, A., 1999. Risk management in the conceptual phase of aproject. International Journal of Project Management 17, 161169.Vidal, L.A., Marle, F., 2012. Asystems thinking approach for project vulnerabilitymanagement. Kybernetes 41, 206228.Ward, S.C., 1999. Assessing and managing important risks. International Journalof Project Management 17, 331336.Ward, S., Chapman, C., 2003. Transformingproject riskmanagement intoprojectuncertaintymanagement.InternationalJournalofProjectManage-ment 21, 97105.Warszawski, A., Sacks, R., 2004. Practical multifactor approach to evaluating riskof investment in engineering projects. Journal of Construction Engineering andManagement 130, 357367.Williams,T.,1993. Risk-managementinfrastructures.InternationalJournalofProject Management 11, 510.Williams, T., 1995. Aclassifiedbibliographyofrecent researchrelatingtoprojectriskmanagement.EuropeanJournalofOperationalResearch85,1838.Williams, T., 1996. The two-dimensionality of project risk. InternationalJournal of Project Management 14, 185186.Winch, G.M., 2003. Managing construction projects. Wiley-Blackwell,Oxford.Wirba, E., Tah, J., Howes, R., 1996. Risk interdependencies and natural languagecomputations. EngineeringConstructionandArchitectural Management 3,251269.Wood, G.D., Ellis, R.C.T., 2003. Riskmanagement practicesofleadingUKcost consultants. Engineering Construction and Architectural Management10, 254262.14 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights from a literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004Xu, T., Tiong, R.L.K., 2001. Risk assessment on contractors' pricing strategies.Construction Management and Economics 19, 7784.Yeo, K., 1990. Risks, classification of estimates, and contingency management.Journal of Management in Engineering 6, 458.Zayed, T., Amer, M., Pan, J., 2008. Assessing risk and uncertainty inherent inChinese highway projects using AHP. International Journal of ProjectManagement 26, 408419.Zeng, J., An, M., Smith, N.J., 2007. Applicationof afuzzybaseddecisionmaking methodology to construction project risk assessment. InternationalJournal of Project Management 25, 589600.Zhang, H., 2007. A redefinition of the project risk process: using vulnerabilitytoopenuptheevent-consequencelink. International Journal of ProjectManagement 25, 694701.Zhang, G., Zou, P.X.W., 2007. Fuzzy analytical hierarchy process riskassessment approach for joint venture construction projects in China.Journal of Construction Engineering and Management 133, 771.Zhi, H., 1995. Riskmanagement foroverseasconstructionprojects. Interna-tional Journal of Project Management 13, 231237.Zou, P.X.W., Zhang, G., Wang, J., 2007. Understanding the key risks in constructionprojects in China. International Journal of Project Management 25, 601614.15 A. Taroun / International Journal of Project Management xx (2013) xxxxxxPlease cite this article as: Taroun, A., Towards a better modelling and assessment of construction risk: Insights froma literature review, International Journal of ProjectManagement (2013), http://dx.doi.org/10.1016/j.ijproman.2013.03.004