dubois gadde 2002

Introduction

The physical substance of a house is a pile of materialsassembled from widely scattered sources. Theyundergo different kinds and degrees of processing inlarge numbers of places, require many types ofhandling over periods that vary greatly in length, anduse the services of a multitude of people organized intomany different sorts of business entity.

These characteristics of the construction industrywere expressed almost 50 years ago in a well knownstudy of the distribution of house building materials(Cox and Goodman, 1956). One of the conclusions ofthe study is that the number of possible permutationsand combinations of speci� c places and entities is enor-mous, even for one product. The complexity of theconstruction operations and the subsequent problemsolving capability needed are perceived as formidable.However, this problem is in fact solved over and overagain as new houses go up in their millions. Similaropinions concerning the complexity of the industryhave been expressed more recently. For example,Shammas-Thoma et al. (1998) discussed ‘all those

remarkable processes that enable the constructionprocess to function at all’. Winch (1987) argued thatconstruction projects are amongst the most complexof all undertakings. Gidado (1996) further emphasizedthis view by stating that there is a continuous increasein the complexity of construction projects.

These underlying conditions shape the industry’sway of functioning and its performance. Now and then� rms in the construction industry are blamed for inef-� ciency of operations (Cox and Thompson, 1997).Speci� cally, it has been argued that a short termperspective promotes suboptimization (Gann, 1996)and hampers innovation and technical development(Dubois and Gadde, 2000). A number of authors haveargued that construction has failed to adopt techniquesthat have improved performance in other industries,such as just-in-time (Low and Mok, 1999), totalquality management (Shammas-Thoma et al., 1998),partnering with suppliers (Cox, 1996), supply chainmanagement (Vrijhoef and Koskela, 2000) and ‘indus-trialization’ of manufacturing processes (Gann, 1996).It seems to be a common view among most authorsthat the construction industry would be better off tochange its behaviour in accordance with the norms ofother industries.

Construction Management and Economics (2002) 20, 621–631

The construction industry as a loosely coupledsystem: implications for productivity and innovation

ANNA DUBOIS* and LARS-ERIK GADDE

Department of Industrial Marketing, Chalmers University of Technology, S-412 96 Gothenburg, Sweden

Received 7 March 2002; accepted 13 June 2002

Previous research suggests that the construction industry is characterized by (1) particular complexity factorsowing to industry speci� c uncertainties and interdependences, and (2) inef� ciency of operations. The aimof this study was to analyse the operations and behaviour of � rms as a means of dealing with complexity.The observations made indicate that the industry as a whole is featured as a loosely coupled system. Takingthis as a starting point, the couplings among activities, resources and actors were analysed in different dimen-sions. The pattern of couplings builds on two interdependent layers: tight couplings in individual projectsand loose couplings based on collective adaptations in the permanent network. It is concluded that thepattern of couplings seems to favour short term productivity while hampering innovation and learning.

Keywords: Construction, loosely coupled system, productivity, innovation

*Author for correspondence. e-mail: [email protected]

Construction Management and EconomicsISSN 0144-6193 print/ISSN 1466-433X online © 2002 Taylor & Francis Ltd

http://www.tandf.co.uk/journalsDOI: 10.1080/01446190210163543

However, assume that Winch (1987) and Gidado(1996) are right in the statements about the particu-larities of construction complexity. If so, it might wellbe that management techniques that improve perfor-mance in other industries are not readily transferableto this context. If construction follows a different logicthen it might even be a mistake to try to adopt manage-ment techniques applied in other contexts.

Aim and scope

The aim of this study was to analyse the operationsand the behaviour of � rms in the construction industry.The perspective chosen holds that the behaviour of� rms can be seen as their attempts to cope with thecomplexity of construction projects. In this respect wesuggest that the industry is to be regarded as a ‘looselycoupled system’ (Weick, 1976).

The construction industry as a looselycoupled system

Complexity in construction

Gidado (1996) argued that complexity in constructionoriginates from a number of sources: resourcesemployed, the environment in which construction takesplace, the level of scienti� c knowledge required, andthe number and interaction of different parts of thework� ow. He distinguished between two main cate-gories of complexity. One is related to ‘uncertainty’and deals with ‘the components that are inherent inthe operation of individual tasks and originate from theresources employed or the environment’. The secondtype of complexity stems from ‘interdependence’among tasks, and represents those sources ofcomplexity that ‘originate from bringing different partstogether to form a work � ow’ (Gidado, 1996).

The uncertainty in the undertaking of individualactivities has four causes: (1) management is unfamiliarwith local resources and the local environment; (2) lackof complete speci� cation for the activities at the con-struction site; (3) lack of uniformity of materials, work,and teams with regard to place and time (every projectis unique); and (4) unpredictability of the environment.Obviously these characteristics make it dif� cult to applya centralized approach to decision-making. Prevailingconditions call for decentralization of authority.

The second determinant of complexity is associatedwith operational interdependence in construction.Gidado (1996) pointed to three factors: (1) the numberof technologies and the interdependence among them;(2) the rigidity of sequence between the various main

operations; and (3) the overlap of stages or elementsof construction. These conditions emanate from twocharacteristics of the industry identi� ed by Eccles(1981). The � rst is ‘the organization of the produc-tion work force into a variety of trades’ and the secondis ‘the practice of subcontracting portions of a projectto special trade contractors by primary contractors’.Both factors cause interdependence, which calls forcoordination. The nature of these interdependencesseems to favour local rather than centralized coordin-ation.

The point of departure of this paper is thatcomplexity emanating from uncertainty and interde-pendence sets the conditions for the behaviour of the� rms in the construction industry. Below, we discusssome central features of the operations in construction.

Central features of construction

It has been argued that construction is inherently a site-speci� c project-based activity (Cox and Thompson,1997). This view is shared by Shirazi et al. (1996), whoconcluded that construction is mainly about coordina-tion of specialized and differentiated tasks at the sitelevel. The emphasis on site-speci� c activities providesus with two central features of construction. The � rstis the focus on individual projects, in terms of decen-tralized decision-making and � nancial control. The pre-vailing organizational arrangements when it comes toresponsibility and authority put the emphasis on theef� ciency of individual projects, which makes sense asa response to the roots of complexity identi� ed above.The strong reliance on localized decision-making isexplained by the fact that management is unfamiliarwith local resources and local environment. The secondfeature is the need for local adjustment at the con-struction site. These adjustments are necessary becauseof the three remaining uncertainty factors: lack of com-plete speci� cation, lack of uniformity and an unpre-dictable environment. When these conditions prevail itis dif� cult (and even unsuitable) to develop compo-nents and systems tailored to the situation at speci� csites. Therefore it is quite unusual for building materialsmanufacturers to develop products adapted to particu-lar contractors or speci� c construction sites. The indus-try still relies on what Stinchcombe (1959) identi� ed asstandardized parts, whereas the use of standardizedactivities tends to be the norm in many other industries.The prevailing uncertainty makes the use of standard-ised parts an appropriate strategy, which is further rein-forced by the bene� ts gained from increasing economiesof scale in the manufacturing of building materials.

The complexity with respect to interdependences alsoseems to favour standardization, and thus local adjust-ments. Owing to the number and interdependences

622 Dubois and Gadde

among technologies, customized solutions from onesupplier would impact on other components and sys-tems. The rigidity of sequences and the overlap of stagesin turn make coordination dif� cult. It is most likely,therefore, that these conditions are better taken care ofthrough decentralization and local adjustments thanthrough centralized activities and customized solutions.

There are some other features of the behaviour ofconstruction � rms that need to be stressed. The strongemphasis on individual projects favours a narrowperspective, both in time and scope. Ef� ciency issupposed to be promoted by competitive tendering.Cox and Thompson (1997) found the perception ofthe actors to be that competitive tendering assures thatsubcontracting is carried out at the lowest possible cost.The strong reliance on competitive tendering explainsthe use of standardized parts. Adapted or customizedsolutions would be dif� cult to use while maintainingthe current tendering procedures. Competitivetendering also sets the conditions for the relationshipamong the parties. Gann (1996) found that the rela-tionships are often typi� ed by market-based, short term interactions between independent businesses. Inaddition, Thompson et al. (1998) identi� ed market-based interaction as the norm for the behaviour, andconcluded that � rms ‘traditionally paid very little atten-tion to the relational elements of business transactions’.The � nal characteristic of behaviour in the industry webring up is the multiple roles of � rms. The activityscope of a � rm tends to be broad, including design,production and distribution in various combinations,which may also vary between different projects. Thedivision of labour among the actors varies greatly fromproject to project and the role of the individual � rmcan be very different (Dubois and Gadde, 2000).

Loose couplings

Our second starting point is to regard the constructionindustry as a loosely coupled system (Weick, 1976).In this section we describe some important aspects ofloose couplings. According to Orton and Weick(1990), any location in an organization contains inter-dependent elements that vary in the number and thestrength of their interdependences. For example, everysingle industrial activity is to some extent interdepen-dent with a number of other activities: they are coupledin various ways. Some of these couplings are ‘tight’while others are ‘loose’. Glassman (1973) discussedthe degree of coupling between two units (events/elements/systems, etc.) on the basis of the activity ofthe variables that the two units share. If two units havefew variables in common, or if the variables in bothare weak compared with other variables in� uencing thetwo units, then they are relatively independent of each

other and thus loosely coupled (Aldrich, 1980).Weick’s (1976) characterization of loose couplings isthat ‘coupled events are responsive but that each eventalso perceives its own identity and some evidence ofits physical or logical separateness’. The attachmentamong the events may be circumscribed, infrequent,weak in its mutual effects, unimportant, and/or slowto respond. Loose couplings may occur in a numberof dimensions: among individuals, among subunits,among organizations, between hierarchical levels,between organizations and environments, among ideas,between activities, and between intentions and actions.

Weick (1976) analysed the potential effects of loosecouplings, which may be functional and/or dysfunc-tional. In this section we direct our attention primarilyto the ways in which loose couplings contribute tohandling complexity in operations. First, a looselycoupled system may be a good system for localizedadaptation where any one element can adjust to andmodify a local unique contingency without affectingthe whole system. Hence, localized adaptations may beswift, relatively economical and substantial. Second,loose couplings serve as buffering mechanisms againstunfavourable conditions in the environment, so theorganization as a whole will not have to respond toeach little change in the environment. As Weick putit: ‘loose couplings allow some parts of an organiza-tion to persist’. Third, loose couplings provide a sensi-tive sensing mechanism. This is a consequence oflocalized adaptation, decentralization and low extentof coordination. It is argued that loosely coupledsystems preserve many independent sensing elementsand therefore know their environments better than istrue for more tightly coupled systems, which have fewerexternally constrained, independent elements. Fourth,loosely coupled systems preserve the identity, unique-ness, and separateness of elements and may thereforegenerate variety. The system can potentially retain agreater number of mutations and novel solutions thanwould be the case with a tightly coupled system. Thegreater freedom in a loosely coupled system wouldimply that the actors deal with problems in a multi-tude of ways, thus favouring variety and innovation.Finally, in a loosely coupled system there is more roomavailable for self-determination by the actors. Accord-ing to Weick, it is likely that a sense of ef� cacy mightbe greater in a loosely coupled system with auton-omous units than it would be in a tightly coupledsystem where discretion is limited.

Table 1 summarizes this discussion of the centralaspects regarding complexity in construction, thecentral features of the industry conduct, and someimportant characteristics of loosely coupled systems. Aloosely coupled system may cope with certain aspectsof the complexity originating from uncertainty and

Construction industry as a loosely coupled system 623

interdependence, since its functions are characterizedby limited central authority and low costs of coordi-nation. The loose coupling, in turn, sets the conditionsfor the behaviour of the � rms in the industry, whichis an important issue for analysis. However, insuggesting that the construction industry features thefunctions of loosely coupled systems we must followOrton and Weick (1990), arguing that the recognitionof an organization ‘being’ a loosely coupled system isthe beginning of the analysis, not the end. Researchersshould not simplify the concept but invoke it: ‘whatelements are loosely coupled, what domains are theycoupled on, and what are the characteristics of thecouplings and decouplings?’ (Orton and Weick, 1990).

Based on observation of the construction industry as‘behaving’ like a loosely coupled system, it is thus fruit-ful to scrutinize the tight and loose couplings prevalentin it. According to Weick (1976), it is the pattern ofcouplings (tight and loose) that together produces theobserved outcomes of a system. In the next section anattempt is made to further explore the pattern of cou-plings and shed more light on the interrelated com-plexity factors and functions of loose couplings.

The pattern of couplings in the constructionnetwork

For an analysis of tight and loose couplings we maybegin with Figure 1, where a construction project isillustrated in its network context. The project may beconsidered as a speci� c temporary network within amore ‘permanent’ network. In Figure 1 � rms A, B andC are all involved in operations at a construction site.Their input in the project comprises resources ofvarious kinds (A1, B1 and C1). The � rms are also

involved in other projects in which they have to coor-dinate their activities and resources with (partly)different sets of other � rms. For example, in Figure 1� rm C needs to consider four different dimensions ofco-ordination: (a) within the individual project (C1with A1 and B1); (b) among � rms involved in supplychains (i.e. with D and E); (c) among differentconstruction projects (C1 with C2); and (d) inter-� rmcoordination beyond the scope of the individual project(i.e. with A and B).

Coordination within construction projects

Owing to (1) the importance of time, (2) the need toperform and coordinate the activities sequentially, and(3) the specialization of actors, there are tight couplingsbetween activities undertaken on site. According toGidado (1996), this is one important factor that makesconstruction complex: ‘. . . in a rigid sequence of work� ow, time or duration change in any specialist’s workmay affect the duration of others or even the overall pro-duction process duration. This sort of knock-on effectmay also affect production cost.’ Furthermore, theactivities are not only sequentially interdependent butalso organized in parallel sequences, i.e. stages or ele-ments of the construction process are overlapping.According to Gidado this adds to the complexity: ‘Theoverlapping of major elements of production is used bypractitioners simply to compress or shorten the produc-tion time. In practice, this process is dictated by a num-ber of resource-dependent factors. Even by consideringthese factors, overlapping may change the interdepen-dence of activities (or trades in particular) within individual elements and also create a new structure ofinterdependences between the roles of the overlappingelements. These changes may increase the effects of


Table 1 Complexity factors, central features of construction and the effects of loose couplings

Complexity in construction

Interdependence Uncertainty Central features of construction Effects of loose coupling

Number of technologies and Lack of complete Focus on single projects Localized adaptationinterdependences activity speci� cation

Local adjustment Buffering mechanismRigidity of sequence Unfamiliarity with between the various main local resources and Utilization of standardized parts Sensing mechanismoperations local environment

Competitive tendering Generation of variationOverlap of stages or Lack of uniformity of elements of construction materials, work and Market-based exchange Self-determination

teams with regard totime and place. Multiple roles

Unpredictability ofenvironment

inherent complexity and uncertainty factors on projectcomplexity.’

Another effect of the strong interdependence amongthe activities undertaken in every construction processis that the consequences of changes are dif� cult toassess and overview. In the ‘Tavistock studies’ it wasfound that: ‘. . .each time a design decision was takenit set in train a chain of consequences which could anddid cause the initial decision to be changed, a clearexample of how decisions and actions depend on oneanother’ (Crichton, 1966, p. 17). These characteris-tics thus require tight couplings between the activitiesundertaken in individual projects.

Coordination within supply chains

Most of the ‘input’ resources used in buildings arestandardized. Furthermore, the chain of activities,including transportation and storage, from the produc-tion of building materials to the site seem to be basedon standardized rules. Typically, large orders are sentdirectly from the factory to the construction site, whilesmaller orders are delivered from the distributors’warehouses. Factory deliveries normally imply ratherlong lead times from order to delivery, whereas thedistributor may be able to deliver on shorter notice.Hence, distributors provide ‘slack’ resources, and thisis important when the exact volumes and timing aredif� cult to foresee, a dif� culty that results from thevery strong interdependence among activities carriedout on site and that may result in delays.

Hence, the couplings in the supply chains in con-struction are both tight and loose. They are loose interms of the coupling between the production of build-ing materials and what is done on site; see e.g. Vrijhoefand Koskela (2000) and Akintoye et al. (2000). The

loose couplings result from the rather long lead timesand the ‘slack’ provided by distributors. The couplingsare tight in the relation between the activities under-taken on site and the activities carried out in the supplychains. If the material has not arrived to the site whenneeded the whole production plan may be jeopardized.

Coordination within � rms

Every � rm involved in on-site activities has to coordi-nate its activities and resources among the differentconstruction projects in which it is involved. Thestrong interdependences among activities performed ateach and every site and the effects of any interdepen-dence in terms of time extensions and delays impliesthat every � rm needs some slack resources. If not, thenthe ‘knock-on’ effects from delays in one project wouldcarry over to other projects. The (� rm) internal coor-dination of activities undertaken and resourcesemployed at different sites may thus even be subjectto competition if the slack is not suf� cient (Crichton,1966). This is of particular importance to � rms special-izing in activities undertaken late in the process.

In addition, the � rms may be specialized in termsof resources, but their roles may vary among projectsand thus also their roles vis-à-vis other � rms involved(Dubois and Gadde, 2000). Gidado (1996) referred tothe learning curve concept, stating that the varyingnature of interdependences or interfaces of roles ofteams in construction may bring about the occurrenceof any one or a number of inherent complexity anduncertainty factors: ‘It is human nature to learn fromexperience and improve in future similar processes;therefore, when roles are repeated over and over bythe same team, it is quite possible that the effect of . . . standard time or cost may decrease’.

Coordination among � rms beyond the individualconstruction project

In construction, one and the same team very seldom(and then rather by coincidence than by consciousplanning) works together in more than one project,and even if some of them meet in another project theirroles vis-à-vis one another may have been altered.Hence, the couplings between activities undertaken atone site and activities undertaken at other sites areloose. Even less tight are the couplings between activ-ities undertaken by different � rms beyond the scopeof an individual construction project.

Gann (1996) argued that dif� culties in creatingcouplings outside individual construction projects havefostered the development of prefabricated standardcomponents, i.e. the types of component that areproduced without prior knowledge of some speci� c


Figure 1 A construction project in its context (modi� edfrom Dubois and Gadde, 2000)

project. This relation probably goes both ways, i.e. theexistence of standard components has made it unnec-essary to develop customized solutions through thecreation of couplings external to construction projects.Regardless of the direction of the relation, on-site, andthus localized, adaptations are very characteristic of theconstruction production system.

The mix of loose and tight couplings

The pattern of couplings in construction is character-ized by tight couplings in individual projects and loosecouplings in the permanent network. The loosecouplings among � rms make it problematic to applythe coordination mechanisms for handling complexitythat are used in other business contexts. Typically, inmost other industries uncertainty and interdependenceare managed through tight couplings among � rms.Relational exchanges and inter-� rm adaptations arecommon means of handling these issues. By contrast,the construction industry is characterized by few inter-� rm adaptations beyond the scope of individualprojects, and � rms tend to rely on short term marketbased exchange. These conditions also imply that theindividuals in the project teams are recombined in eachproject, which further complicates coordination.

Considering the small number of � rm-speci� c adap-tations it is quite surprising that it is possible to createthe tight couplings in the projects. The explanation isthat the construction industry is characterized bycertain ‘collective adaptations’ (Dubois and Gadde,2000). For example, the standardized components andsystems used have been developed through continuouscollective efforts among material producers, contrac-tors and the governmental authorities who prescribenorms and other conditions. Collective adaptations areformed in what can be identi� ed as ‘a community ofpractice’ (Brown and Duguid, 1998). These authorsargued that a great deal of knowledge is both preservedand held collectively. Collective knowledge is gener-ated when people work together in tightly knit groupsknown as ‘communities of practice’. The claim is thatthis type of common practice promotes collectiveknowledge, shared sense-making and distributedunderstanding. A community of practice develops ashared understanding of what is done and how it isdone. In this way a strong community of practicereduces uncertainty and serves as an informal coordi-nation mechanism in loosely coupled systems. Forexample, Meyer (1975) argued that the school systemin the USA works because everyone else knows roughlywhat is going on. The community of practice forms acommon culture, which functions as a template forhow � rms perceive the environment. It also serves asa pattern for action and guides the behaviour of � rms.

Powell (1991) discussed the bene� ts of this type ofshared expectations: ‘Shared expectations arise thatprovide psychological security, reduce the cost of infor-mation processing, and facilitates the coordination ofdifferent activities. Moreover, established conceptionsof the way things are done can be very bene� cial;members of an organization can use these stable expec-tations to predict the behaviour of others.’

The construction industry relies on a strong com-munity of practice. Important aspects of this commonpractice are revealed in a study by Kadefors (1995).Governmental regulations have a signi� cant impact onthe design and construction of houses. Building codes,norms, and principles for housing subsidies imposerequirements favouring certain standards. The regula-tory system concerning the work environment andworkers’ health and safety contributes to reinforcing thecommunity of practice. The industry itself is also asource of formal standardization. The � rms involvedhave established numerous forms of common contractformulas, which set standards in terms of operations,components, documentation, and work principles.Moreover, the tendering procedure requires that sup-pliers’ offerings are standardized. Without standardiza-tion, contractors would not be able to evaluate thedifferent offerings. The generic roles of the participantsin the processes of design, planning and constructionare also standardized. Individual � rms take on differentroles in different projects. Therefore, the generic rolesof designers, general contractors and subcontractors(plumbers, carpenters, etc.) must be similar in differentprojects. These roles are closely related to standardiza-tion of skills and knowledge, which follows from theexistence of an informal control system. This isacknowledged in standard contracts, according towhich the quality of the contractor’s work should con-form to ‘current standards of workmanship’ (Kadefors,1995). The need for formulations of this type stemsfrom the dif� culties in exactly specifying every detail ofeach task in the contract. Reliance on the standard ofworkmanship helps reduce the type of uncertaintydescribed by Gidado (1996) as ‘lack of complete spec-i� cation’. These conditions are important reasons forthe strong reliance on decentralization of authority andthe requirements for localized adaptations.

Discussion

The aim of this study was to understand the logic of theoperations in the construction industry, and it hasrevealed that the behaviour of the � rms differs consider-ably from what is common in other industries, particu-larly in terms of the absence of inter-� rm adaptat-ions. The industry operates similarly to what have been


identi� ed as loosely coupled systems (Weick, 1976).The pattern of tight and loose couplings can be inter-preted as a means of coping with the prevailing com-plexity in construction operations. The tight couplingsin individual projects combined with the loose couplingsin the permanent network embedded in the communityof practice make it possible to come to grips with uncer-tainty and interdependence. In particular, it appearsthat the loose couplings in the permanent network pro-vide the slack necessary to handle the tight couplings inprojects. Our discussion of the features of industry oper-ations revealed six central aspects. The focus on indi-vidual projects, the use of standardized components,local adjustments, and the multiple roles played by � rmsallow both for handling complexity in individual pro-jects and securing economies of scale in manufacturing.Our overall conclusion is thus that the behaviour of theindustry seems to be an appropriate response to theinherent complexity of construction projects.

However, the central features identi� ed also includecompetitive tendering and market-based exchangeamong � rms. It is not clear whether loose couplings inthese respects are necessary to attain the bene� ts ofslack and � exibility. Therefore, a further exploration ofthe relationship between complexity and the nature ofbusiness conditions and exchange is the � rst topic ofthe discussion. Our � ndings indicate that the prevailingpattern of couplings seems favourable for the short termproductivity of individual projects, whereas the longterm effects are less obvious. The second issue for dis-cussion is therefore the consequences implied by thesecouplings for innovation and learning in the permanentnetwork. Finally, we introduce some potential featuresemanating from alternative patterns of couplings.

Complexity and the nature of exchange

Crichton (1966) found construction to be character-ized by technical interdependence and organizationalindependence. The organizational arrangements in theindustry are based on the assumption that dependenceon individual counterparts should be avoided, becausedependence might impose problems in variousrespects. This view used to be current in other indus-tries as well, but it has gradually been abandonedthrough recognition of the advantages that can begained from close relationships. Obtaining these bene-� ts entails counterpart speci� c adjustments that, inturn, necessitates dependence on speci� c partners(Gadde and Håkansson, 2001). Therefore, it is mostprobable that development of close relationships in thepermanent network would improve performance inconstruction as well.

Shammas-Toma et al. (1998) argued that thetendering system and the short term perspective are to

blame for many shortcomings in construction, forexample the problems of adopting concurrent engi-neering practices and the dif� culties in integratingdesign and building activities. These problems emanatemainly from the sequence of operations in the opentender form of the building process. Because of thisprocedure, design affects construction planning whileconstruction planning cannot affect design. Shammas-Toma et al. (1998) illustrated the consequences forcontractors who have to build according to speci� eddimensions, shapes, strength requirements, etc.,regardless of the problems that the design speci� cationmay pose during construction. Similar criticism wasexpressed by Vrijhoef et al. (2001), who argued thatthe actors involved in the design project organizationhave no common and clear understanding of whatshould be designed. Thus, in this respect relationalexchange could contribute to improved coordinationand reduce complexity stemming from interdepen-dence. The other dimension of complexity is concernedwith uncertainty. Competitive tendering and market-based exchange reduce the uncertainty associated withthe evaluation of offerings and switching costs. On theother hand, interaction in close relationships can beused as means for reducing other types of uncertainty,for example need uncertainty and transaction uncer-tainty (Ford et al., 1998).

The implication of this discussion is that changes inthe pattern of couplings may affect both performanceand complexity. It seems likely that other patternscould improve the performance in construction withoutincreasing the complexity. As argued above, four of thecentral features of the industry’s behaviour seem to berelevant means for managing complexity. However,when it comes to competitive tendering and market-based exchange the situation is different. The analysisleads us to question whether these conditions arenecessary for gaining the bene� ts from local adjust-ment, standardized components, focus on individualprojects, and the multiple roles of the � rms. Tightercouplings among � rms might be bene� cial to theoverall performance in construction. There also seemsto be increasing interest among � rms in developingcloser relationships. For example, Cox and Thompson(1997) stated that the search for more collaborativecontractual relations has become a contemporarytheme in the construction industry. However, theauthors (and others) found that these efforts have notyet been very successful. Our conclusion is that achange in this direction must be dif� cult to undertake,because it is not in accordance with the norms of thecommunity of practice. We agree with Kornelius andWarmelink (1998), who argued that coordinationthrough bilateral relationships is problematic inconstruction.


The organizational independence identi� ed byCrichton (1966) is not only a characteristic of inter-� rmrelationships. The decentralization of authority to theindividual project leads to loose couplings even betweendifferent entities within � rms. We have argued thatthese conditions provide opportunities for localizedadaptations and support self-determination. On theother hand, large contractors have an obvious interestin taking advantage of potential economies of scale inpurchasing. Decentralization of authority might con-strain these efforts, because in loosely coupled systemsa centrally located authority has limited possibilities tointervene in local operations. According to Weick(1976), the same mechanisms that work as buffers byisolating trouble spots, and thus preventing the troublefrom spreading, also make it dif� cult to repair the defec-tive element. These conditions may, however, be rep-resentative of other project based activities as well. Forinstance, von Krogh (1998) observed similar tendenciesin R&D projects. O’Dell and Grayson (1998) arguedthat decentralised decision making in temporary organ-isations makes project leaders focus on maximizingtheir own accomplishments and rewards. Therefore,they might act in ways that contradict the goals of theorganization as a whole. The authors conclude that toomuch emphasis on the individual project’s self-deter-mination results in situations where the left hand doesnot know what the right hand is doing.

Loose couplings and innovation

Thus far we have dealt with the effects of the pattern ofcouplings in terms of ef� ciency and productivity. Below,we focus on some of the consequences for innovationand dynamics. According to Teece (1998) the opportu-nities for learning are closely related to previous activi-ties and experiences: if many aspects of a � rm’s learningenvironment change simultaneously the ability to formcognitive structures favouring learning becomes severelyrestricted. This is a problem because learning is aprocess of trial, feedback and evaluation. Gann (1996)argued that this process is seldom successful in con-struction, and concluded that ‘each house is treated as apilot model for a design that never had any runs’. Itseems to be the case that the pattern of couplings doesnot foster economies of scale in design, planning, andconstruction, although they are bene� cial for economiesin manufacturing of building materials.

On the other hand, these industry conditions shouldbe favourable for the development of new ideas. Thepattern of couplings makes each construction site anexperimental workshop. In complex networks experi-mentation is an important breeding ground for innova-tion (Gadde and Håkansson, 2001). One typical

outcome of loose couplings is the ability to generate vari-ation (Weick, 1976). Localized adaptations imply thatany one element can adjust to local contingencies. Thismeans that loosely coupled systems potentially can gen-erate a larger number of mutations and novel solutionsthan would be the case with a tightly coupled system,because the actors deal with problems in a multitude ofways. However, Weick argued that, although a local setof elements can adapt to local idiosyncrasies withoutinvolving the whole system, this same loose couplingcould also forestall the spread of advantageous muta-tions that exist somewhere in the system. Hence, theloosely coupled system may contain novel solutions tonew problems, but the very structure that allows thesemutations to � ourish may prevent their diffusion. Theseconditions prevail in construction and may be explainedby the pattern of couplings. Below we discuss explana-tions related to the project, the individual � rm and therelationships among the actors.

First, the project organization does not promotelearning. One reason is the temporary nature of the pro-ject, offering no guarantee of further contacts amongteam members. The consequences were discussed byCrichton (1966, p. 22): ‘. . . there is no input of com-monly shared experience of other building processes.Each member of the building team brings little morethan his own accumulated experiences (and prejudices)to bear on current problems. Learning, in the sense ofadaptations brought about by experience, is therefore aslow and uncertain process that takes place at an indi-vidual level rather than at industry level.’

However, time limitations also make individuallearning problematic. For example, von Krogh (1998)observed that time constraints made it dif� cult for indi-viduals to get the most learning bene� t out of R&Dprojects. He also argued that too little effort is devotedto transmitting knowledge and experience from oneproject to another. Projects are problematic in thisrespect because they do not have an organizationalmemory (Björkegren, 1998). They lack the naturaltransfer mechanisms of permanent organizations wherestructures and routines can contribute to knowledgeabsorption. Therefore learning needs to be transferredvia the level of the � rm.

The second explanation of the problems with innova-tion in construction relates to the organizationalarrangements within the � rm. In this respect loose cou-plings not only make it dif� cult to intervene in localizeddecision-making, they also prohibit learning and inno-vation because in strongly decentralized structures theleft hand not only does not know what the right hand isdoing but also it may not even know that there is a righthand (O’Dell and Grayson, 1998). Therefore, in anorganization based mainly on decentralization and pro-ject activities ‘lies unknown a vast treasure house of


knowledge, know-how and best practice’ (O’Dell andGrayson, 1998). These conditions are prevalent in con-struction as well. The activities at construction sites gen-erate a lot of ideas from creative problem-solving tasks.However, the pattern of couplings in the industry is anobstacle to their diffusion.

Third, the loose couplings in the permanent networkserve as a barrier to innovation. Long-term relationshipsand adaptations beyond individual construction projectsare necessary to foster learning and innovation. Forexample, Loasby (1976) argued that learning cannottake place through anonymous contracting but requirescontinuous interaction through which an individual or acompany commits to the group and thus becomes onewithin the group. The existing market based short termexchanges cause problems in this respect. The outcomeof this procedure is that the constellation of � rmsinvolved in the temporary network does not have jointplans beyond the project (Thompson et al., 1998).Therefore, neither the individual nor the companybecomes ‘one within the group’. They become ‘onewithin a group’ the constitution of which changes fromone project to the next. The problems associated withthese organizational arrangements are analysed by sev-eral authors. Kreiner (1995) pointed out the danger ofthe short term based project focus, arguing that the factthat projects occupy only a bracket in time, and thus haveneither history nor future, allows evolutionary processeslittle scope for improving performance’. Cox andThompson (1997) argued that because the constellationof actors is continually changing it is dif� cult to make useof experience gained in previous projects. They con-cluded that this creates particular cost inef� ciencies forthe client as a new learning curve is climbed each time.

Tighter couplings among � rms in the permanentnetwork could thus enhance the opportunities for inno-vation. We have argued above that such conditionsmight also improve the possibilities of reducing uncer-tainty, through the continuous interaction in close rela-tionships. Furthermore, if couplings become tighter itis most likely that the parties will � nd new ways toadapt to each other, as has been important for inno-vation in other industries. For example, some of theadjustments now undertaken at the construction sitemight be conducted more ef� ciently up-stream of thesupply chain through utilization of more specializedresources in terms of machinery and manpower (e.g.Vrijhoef and Koskela, 2000).

The fourth barrier to innovation is found in thestrong community of practice. We identi� ed the com-munity of practice as a means of enhancing productiv-ity and ef� ciency, because it allowed for tight projectcouplings in spite of the loose couplings in the perma-nent network. The community of practice stabilizesconditions that promote short term productivity.

However, the same conditions hamper innovationbecause they tend to make � rms similar and indepen-dent. This is a problem where learning is concerned,because heterogeneity and interdependence are greaterspurs to collaborative action than homogeneity and dis-cipline (Powell, 1998). In construction the resources ofdifferent suppliers are quite homogeneous and a con-tractor could not expect to learn more from one of themthan from another. These conditions differ from the sit-uation in many other industries (Powell, 1998).Furthermore, government regulations and industrystandards make the system dif� cult to change, and thisin turn hampers innovation. According to Kadefors(1995), the existence of joint industry standards sim-pli� es work considerably. However, these standardsalso imply that only certain well tested constructions areincluded, and therefore the technical solutions andwork procedures actually are reduced. The tenderingsystem favouring standard offerings thus functions as abarrier to innovation and the creation of new solutions.

Alternative patterns of couplings

We argue here that the construction industry has thefeatures of a loosely coupled system. The particularpattern of couplings favours productivity in projects,while innovation suffers. These characteristics havemade the industry as a whole lag behind other indus-tries in terms of traditional performance measures.Depending on what theoretical foundations have beenbuilt on, this observation has led researchers andconsultants to prescribe either more competition ormore cooperation to increase the performance of theindustry as a whole.

The strong project focus makes coordination in otherdimensions dif� cult or even pointless. Each project isconsidered to have a life of its own – without eitherhistory or future. Furthermore, it is only looselycoupled to the overall network structure, thus havingfew connections with other projects. Therefore, perfor-mance criteria relate to what takes place within theboundary of the individual project. This focus makesit problematic for a contractor to coordinate efforts indifferent projects. Furthermore, it complicates inter-� rm cooperation. The boundary around individualprojects calls for standardized interfaces among � rms,favouring short term productivity and hinderinglearning. Strongly focusing on one single dimension ofperformance means neglecting others.

In construction, the most obvious experiment wouldbe to put less emphasis on the project boundary. Such achange would allow for increasing coordination in otherdimensions. Successful experiments of this kind can beobserved in other industries. For example, ‘just-in-time’


delivery is the outcome of close coordination of supplychains, and ‘customization’ is the outcome of close col-laboration in inter-� rm development teams. Such net-work structures emphasize other performance criteria,and are based on other combinations of tight and loosecouplings. Vrijhoef and Koskela (2000) suggested anumber of ways in which an increasing supply chain ori-entation might contribute to improved performance.Among other things, they advocated changes in theinterface between site activities and the supply chain andtransfer of activities from the site to the supply chain.

What these alternative efforts have in common is� rst that they are based on inter-� rm cooperation andcounterpart-speci� c adjustments leading to interactiveeffects. Second, connections between relationshipsmake it possible to build on previous interactive effectswhich, in turn, foster learning in the structure as awhole. The main characteristic of these successfulattempts is the interdependence among organizationsand projects, in contrast to the independence typicalof projects and � rms in construction.

However, when we suggest more attention to inter-� rm cooperation, we must also emphasize that thedialectic nature of couplings stressed by Weick (1976)should not be forgotten. Couplings are interrelated, andthus any change of a coupling impacts on the others.The pattern of couplings in the construction industryfavouring project ef� ciency is clearly an obstacle toinnovation and learning. We have pointed out somepossible modi� cations of the present pattern of cou-plings. However, following Weick (1976), changingsome of the couplings necessarily means that other cou-plings will change as well. It is the pattern of couplingsthat shapes (and is affected by) the behaviour of theactors. Different patterns have different consequencesfor complexity. Each pattern reduces some uncertain-ties and increases others, in the same way that it solvessome interdependences and creates new ones.

References

Akintoye, A., McIntosh, G. and Fitzgerald, E. (2000) Asurvey of supply chain collaboration and management inthe UK construction industry. European Journal of Pur-chasing & Supply Management, 6, 159–68.

Aldrich, E. (1980) Organizations and Environments, Prentice-Hall, Englewood Cliffs, NJ.

Björkegren, C. (1998) Learning for the next project, inEasterby-Smith, M., Araujo, L. and Burgoyne, J. (eds),Proceedings of the 3rd Conference on Organizational Learning,Lancaster University, pp. 107–23.

Brown, J. and Duguid, P. (1998) Organizational knowledge.California Management Review, 40(3), 90–111.

Cox, A. (1996) Relational competence and strategic procure-ment management. European Journal of Purchasing andSupply Management, 2(1), 57–70.

Cox, A. and Thompson, I. (1997) ‘Fit for purpose’ contrac-tual relations: determining a theoretical framework forconstruction projects. European Journal of Purchasing andSupply Management, 3, 127–35.

Cox, R. and Goodman, C. (1956) Marketing of house-building materials. Journal of Marketing, 21(1), 36–61.

Crichton, C. (1966) Interdependence and Uncertainty: A Studyof the Building Industry, Tavistock, London.

Dubois, A. and Gadde, L.-E. (2000) Supply strategy andnetwork effects: purchasing behaviour in the constructionindustry. European Journal of Purchasing and Supply Man-agement, 6, 207–15.

Eccles, R. (1981) Bureaucratic versus craft administration:the relationship of market structure to the construction� rm. Administrative Science Quarterly, 26, 449–69.

Ford, D., Gadde, L. E., Håkansson, H., Lundgren, A.,Snehota, I., Wilson, D. and Turnbull, P. (1998) ManagingBusiness Relationships, Wiley, Chichester.

Gadde, L. E. and Håkansson, H. (2001) Supply NetworkStrategies, Wiley, Chichester.

Gann, D. (1996) Construction as a manufacturing process?Similarities and differences between industrialized housingand car production in Japan. Construction Management andEconomics, 14, 437–50.

Gidado, K. I. (1996) Project complexity: the focal point ofconstruction production planning. Construction Manage-ment and Economics, 14, 213–25.

Glassman, R. (1973) Persistence and loose coupling in livingsystems. Behavioural Science, 18, 83–98.

Kadefors, A. (1995) Institutions in building projects: impli-cations for � exibility and change. Scandinavian Journal ofManagement, 11(4), 395–408.

Kornelius, L. and Warmelink, J. W. F. (1998) The virtualcorporation: learning from construction. Supply ChainManagement, 3(4), 193–202.

Kreiner, K. (1995) In search of relevance: project manage-ment in drifting environments. Scandinavian Journal ofManagement, 11(4), 335–46.

Loasby, B. J. (1976) Choice, Complexity and Ignorance,Cambridge University Press.

Low, S. P. and Mok, S. H. (1999) The application of JITphilosophy to construction: a case study in site layout.Construction Management and Economics, 17, 657–68.

Meyer, J. (1975) Notes on a structure of educational orga-nizations; quoted in Weick (1976).

O’Dell, C. and Grayson, J. (1998) If only we knew what weknow: identi� cation and transfer of international best prac-tices. California Management Review, 40(3), 154–74.

Orton, J. D. and Weick, K. E. (1990) Loosely coupledsystems: a reconceptualization. Academy of ManagementReview, 15(2), 203–23.

Powell, W. (1991) Neither market nor hierarchy: networkforms of organization, in Thompson, G., Frances, J.,Levacic, R. and Mitchell, J. (eds), Markets, Hierarchies &Networks, Sage, London, pp. 265–76.

Powell, W. (1998) Learning from collaboration: knowledgeand networks in the biotechnology and pharmaceuticalindustries. California Management Review, 40(3), 228–40

Shammas-Toma, M., Seymour, D. and Clark, L. (1998)Obstacles to implementing total quality management in


the UK construction industry. Construction Managementand Economics, 16, 177–92.

Shirazi, B., Langford, D. and Rowlinson, S. (1996) Organi-zational structures in the construction industry. Construc-tion Management and Economics, 14(3), 199–212.

Stinchcombe, A. (1959) Bureaucratic and craft administrationof production. Administrative Science Quarterly, 4, 168–87.

Teece, D. (1998) Capturing value from knowledge assets:the new economy, markets for know-how, and intangibleassets. California Management Review, 40(3), 55–79.

Thompson, I., Cox, A. and Anderson, L. (1998) Contractingstrategies for the project environment. European Journal ofPurchasing and Supply Management, 4, 31–41.

von Krogh, G. (1998) Care in knowledge creation. CaliforniaManagement Review, 40(3), 133–53.

Vrijhoef, R. and Koskela, L. (2000) The four roles of supplychain management in construction. European Journal ofPurchasing & Supply Management, 6, 169–78).

Vrijhoef, R., Koskela, L. and Howell, G. (2001)Understanding construction supply chains: an alternativeinterpretation, in Chua, D. and Ballard, G. (eds) Proceedingsof the 9th International Group for Lean Construction Conference,National University of Singapore, pp. 185–98.

Weick, K. E. (1976) Educational organizations as looselycoupled systems. Administrative Science Quarterly, 21, 1–19.

Winch, G. (1987) The construction � rm and the construc-tion process: the allocation of resources to the construc-tion project. In Lansley, P. and Harlow, P (eds), ManagingConstruction Worldwide, Vol. 2, Productivity and Humanfactors, E.& F. N. Spon, London.


dubois gadde 2002

Documents