summery
Impact of Team Integration and Group Cohesion on Project Delivery Performance
Bryan Franz, Ph.D., A.M.ASCE1; Robert Leicht, Ph.D., A.M.ASCE2; Keith Molenaar, Ph.D.3; and John Messner, Ph.D.4
Abstract: The architecture, engineering, and construction (AEC) industry is often criticized for its fragmented approach to project delivery. Traditional procurement and contracting intentionally serves to isolate designers from contractors to provide checks and balances, but limits opportunities for collaboration. This research presents a structural modeling approach to studying the role of integration in the performance of building construction projects. A sample data set of 204 completed projects was collected to compare cost, schedule, and quality performance under different delivery methods. Integration of project teams was proposed and tested in the form of two latent constructs—team integration and group cohesion—that mediate the link between delivery methods and performance. More integrated teams interacted with more participants from all levels of the building construction process, from designers to specialty trade contractors. These interactions included design charrettes, joint goal setting, and multidisciplinary building information modeling (BIM) uses. The selected project delivery method had a significant effect on team integration. Delivery methods that involved the builder and specialty trade contractors before schematic design achieved higher levels of integration and were more equipped to control project schedule growth. Cohesive teams were characterized by better chemistry, goal commitment, and timeliness of communication. Project delivery methods that included cost transparency with open-book contracts and qualification-based selection of the builder resulted in more cohesive teams and a lower average project cost growth. Additionally, the owner’s perception of their turnover experience and building system quality was rated higher for cohesive teams. Under- standing how delivery decisions influence the integration and development of their project teams will make building owners more aware of how those decisions ultimately affect the project’s performance. DOI: 10.1061/(ASCE)CO.1943-7862.0001219. © 2016 American Society of Civil Engineers.
Author keywords: Contracting; Integrated project delivery (IPD); Structural equation modeling (SEM); Collaboration.
Introduction
There is a growing consensus in the architecture, engineering, and construction (AEC) industry that integration of people and proc- esses is an effective means of improving project performance (Walker and Hampson 2003; Smyth and Pryke 2008). Authors also suggest that some project delivery methods are more integrated than others. Building on Konchar and Sanvido’s (1998) seminal work, follow-on studies found that design-build (DB) and construc- tion manager at risk (CMR) yield higher levels of integration than a traditional design-bid-build (DBB) approach (Mollaoglu-Korkmaz et al. 2013). As a result, some owners are considering alternative delivery options, including integrated project delivery (IPD), that promote the integration of design and construction disciplines.
Success stories from IPD projects are seeing publication in trade magazines and empirical performance comparisons are beginning to appear in the literature (El Asmar et al. 2013). Despite these reported successes, owners, architects, and contractors have reser- vations about IPD as a delivery method that are limiting its wide- spread adoption (Kent and Becerik-Gerber 2010). For many, the benefits of integration are clear—improved teamwork, better com- munication, and reliable coordination to name a few. However, it is less clear which project delivery methods are more conducive to integration and to what extent integration affects performance.
Recognizing this gap in knowledge, the authors attempt to model the concept of integration to better understand its role in project performance. Adopting Baiden and Price’s (2011) concep- tualization of an integrated team in construction, the authors define integration as the “merging of different disciplines or organizations into a single cohesive and mutually supporting unit, with alignment of processes and cultures.” This definition weaves together two dis- tinct streams of research relating to integration that are examined separately in this study: (1) a focus on interfirm interactions and (2) the development of a common culture. Labeled team integra- tion in this study, the former refers to the degree to which organ- izations engage in interfirm interactions. Construction projects are a dynamic social network of interdependent, yet contractually dis- jointed organizations (Chinowsky et al. 2010). Many authors (Love et al. 1998; Moore and Dainty 2001; Bromley et al. 2003) view the removal or diminishing of formal organizational boundaries in these project networks as a necessary step toward integration. With fewer boundaries, team members are theorized to be more collabo- rative and willing to interact with individuals outside their own organization. The latter stream of integration research, referred to
1Assistant Professor, M.E. Rinker, Senior School of Construction Management, Univ. of Florida, 573 Newell Dr., Gainesville, FL 32603 (corresponding author). E-mail: bfranz@ufl.edu
2Associate Professor, Dept. of Architectural Engineering, Pennsylvania State Univ., 104 Engineering Unit A, University Park, PA 16801. E-mail: rmleicht@engr.psu.edu
3Professor, Construction Engineering and Management, Univ. of Colorado Boulder, ECOT 444, Boulder, CO 80309. E-mail: keith .molenaar@colorado.edu
4Professor, Dept. of Architectural Engineering, Pennsylvania State Univ., 104 Engineering Unit A, University Park, PA 16801. E-mail: jmessner@engr.psu.edu
Note. This manuscript was submitted on January 21, 2016; approved on June 24, 2016; published online on August 5, 2016. Discussion period open until January 5, 2017; separate discussions must be submitted for individual papers. This paper is part of the Journal of Construction Engineering and Management, © ASCE, ISSN 0733-9364.
© ASCE 04016088-1 J. Constr. Eng. Manage.
J. Constr. Eng. Manage., 04016088
D ow
nl oa
de d
fr om
a sc
el ib
ra ry
.o rg
b y
C ar
ne gi
e M
el on
U ni
ve rs
it y
on 1
0/ 04
/1 6.
C op
yr ig
ht A
S C
E . F
or p
er so
na l
us e
on ly
; al
l ri
gh ts
r es
er ve
d.
as group cohesion, studies the degree to which the project team develops into a single unit with a common culture. The creation of a unique identity is an important step for new teams to achieve integration (Baiden et al. 2006). That identity is born when indi- viduals put aside the differing goals and schemas from their parent organizations and, instead, develop a common culture alongside other members of the project team. By developing a common cul- ture, teams are believed to have a greater commitment and collec- tive understanding of project goals (Moore and Dainty 2001; Mollaoglu-Korkmaz et al. 2013), respect for one another (Dainty et al. 2001), trust (Kumaraswamy et al. 2005), and improved flow of communication (Evbuomwana and Anumba 1998).
Consistent with theories that working relationships determine the effectiveness of a project’s execution (Smyth and Pryke 2008), the authors hypothesize that team integration and group cohesion mediate the effects of project delivery methods on project perfor- mance, as shown as a conceptual model in Fig. 1. The conceptual model identifies multiple pathways, both direct and indirect. With a better understanding of the mechanisms of action through which project delivery affects performance, the authors attempt to advance both theoretical and empirical research on integration in the AEC industry. The model presented in this research helps us to under- stand which project delivery methods influence the development of integrated and cohesive teams and the extent of their contribution to a successful project.
This paper is structured as follows. Building upon the concep- tual model in Fig. 1, each construct is developed through the liter- ature and a set of hypotheses is proposed to be tested using structural equation modeling (SEM). Then, the data collection methodology, the measurement variables, and the analysis methods are presented. Lastly, the results of the SEM are discussed along- side limitations and practical applications of the findings.
Hypotheses
Current project delivery literature does not give enough attention to the underlying factors that may explain performance differences across methods. Expanding on our understanding that project de- livery methods directly influence project performance, this research proposes two constructs as mediators of that relationship: team in- tegration and group cohesion. From the conceptual model in Fig. 1, three hypotheses are proposed to examine the direct and indirect influences on project performance.
Direct Influences on Project Performance
Hypothesis 1: Greater team integration has a positive effect on project performance and group cohesion.
Hypothesis 2: Higher levels of group cohesion have a positive effect on project performance. In our study, project performance refers to the cost, schedule, and quality outcomes reported after final completion. Traditionally, project performance is studied in two ways—qualitatively, to explore the effect of new approaches, processes, and techniques within the context of a case study project, and quantitatively, to confirm the effect of multiple factors across larger samples. Previ- ously cited qualitative research identifies the benefits of integration, but there is little agreement on whether those benefits are realized as improvements in cost, schedule, or quality performance. By comparison, the direct performance relationship between delivery methods and performance is well-researched. Many recent quanti- tative studies focus on the direct effect that the delivery method (Hale et al. 2009; Korkmaz et al. 2010; El Asmar et al. 2013), contract payment terms (Chan et al. 2007; Bogus et al. 2010), and procurement procedures (El Wardani et al. 2006) have on project performance. In other words, the theory that certain delivery methods can improve performance is well developed, but the how and why are only explored to a lesser extent. When taken together, Hypotheses 1 and 2 form the basis of most theories on integration in the AEC industry. Since this research is the first to model team integration and group cohesion in the context of project delivery, the proposed hypotheses make no attempt to relate latent constructs with specific outcome metrics.
Indirect Influences on Project Performance
Hypothesis 3: Project delivery methods have an effect on team integration, group cohesion, and project performance. Previous studies demonstrate that project delivery methods with earlier contractor involvement in the design process consistently outperform the alternatives (Hale et al. 2009; Korkmaz et al. 2010; El Asmar et al. 2013). Hypothesis 3 will test if this direct relationship remains true in the presence of multiple indirect in- fluences, specifically through the latent constructs of team inte- gration and group cohesion. This hypothesis is motivated by the theory that the structure of teams has a significant role in their development and their execution of the project (Smyth and Pryke 2008). This research views the delivery method as a primary con- tributor to the interaction and behavior of teams on construction projects.
Methodology
The goal of this research was to test the explanatory value of integration as a mediator in explaining the influence of project delivery methods on project performance. To achieve this goal, our research approach involved the collection of empirical data from completed building projects for use in statistical modeling. An advisory board of 12 industry practitioners was assembled to guide the research. This advisory board was composed of two construction managers, two specialty contractors, three owners (two private and one public), two lawyers, and one architect, with each member having at least 10 years of experience working in the construction industry. Members were selected to represent the interests from major professional organizations, such as the Design-Build Institute of America (DBIA), the Construction Management Association of America (CMAA), the Associated General Contractors of America (AGC), and the American Insti- tute of Architects (AIA), among others. The advisory board was not intended to represent the voice of an entire industry, but rather as a group interested in understanding and improving the delivery of building construction projects.
Fig. 1. Conceptual model
© ASCE 04016088-2 J. Constr. Eng. Manage.
J. Constr. Eng. Manage., 04016088
D ow
nl oa
de d
fr om
a sc
el ib
ra ry
.o rg
b y
C ar
ne gi
e M
el on
U ni
ve rs
it y
on 1
0/ 04
/1 6.
C op
yr ig
ht A
S C
E . F
or p
er so
na l
us e
on ly
; al
l ri
gh ts
r es
er ve
d.
Development of the Project Survey Questionnaire
The data collection instrument for this research was a survey ques- tionnaire. The unit of analysis was a completed building construc- tion project. The questionnaire was developed through a thorough review of past literature and active cooperation from the advisory board during a two-day structured research charrette. The research charrette provided an opportunity for the advisory board members to interact with one another, under the guidance of senior research- ers, to identify the behaviors or practices found in successful project teams. These practices were then translated into draft survey questions. When combined with prior survey questions from the literature, such as Konchar and Sanvido’s (1998) study, the initial version of the questionnaire was over 200 questions in length. The scope of the survey was then refined by removing redundant ques- tions and having the advisory board identify which data were most critical to the study. Next, the questionnaire was tested on 14 pilot projects. This pilot testing verified that respondents could correctly interpret the questions being asked and resulted in several changes to the wording of specific questions. Additionally, the research team identified the need for a list of frequently asked questions (FAQs) to accompany the survey.
Data Collection Procedures
After developing and piloting the data collection instrument, the survey was distributed to the study’s targeted population. The scope of the study was limited to new, commercial building projects that achieved final completion between 2008 and 2013. Since a truly simple random sample of recent construction projects was not feasible, this research obtained a large-scale convenience sample. Mailing lists from professional organizations were used to reach a diverse subset of the construction industry. Respondents were asked to fill out the survey using information from their most re- cently completed project to reduce any bias toward choosing only high-performing projects. The survey was intended primarily for owners, as the party with the most complete knowledge of the project; however, contractors, construction managers, and design- ers were not discouraged from participating. Surveys were sent as a PDF form for e-mail distributions and as a double-sided paper version for postal mailings. Both forms of distribution were accom- panied by a cover letter and list of FAQs that explained the purpose of the study and listed definitions for several key terms. Respond- ents that received the PDF form had the option of printing the sur- vey and answering by hand or completing the form electronically.
A common challenge in project delivery research is obtaining a representative sample of construction projects. Approximately 2,500 surveys were sent via postal mail to graduates of the Archi- tectural Engineering Department at Penn State and the Construc- tion, Engineering, and Management program at the University of Colorado Boulder. A total of 41 of these surveys were returned, representing a 1.6% response rate. Nearly 6,000 surveys were distributed by e-mail to mailing lists for the Construction Owners Association of America (COAA), the Design-Build Institute of America (DBIA), the Construction Management Association of America (CMAA), the Association for the Advancement of Cost Engineering (AACEI), and the Partnership for Achieving Construc- tion Excellence (PACE) at Penn State. The research team received 290 returned surveys from the e-mail distribution, representing a 4.8% response rate. When combined with postal mail responses, a total of 331 surveys were received for this study, resulting in a 3.9% overall response rate. Of these responses, a total of 204 projects were both within the scope of the study and successfully verified for accuracy, thus forming the sample data set. The demographics of these projects are discussed later in the paper.
Verification of Data and Preparation for Analysis
The project information in each returned survey was verified for accuracy before analysis. Immediately upon receipt, returned sur- veys were reviewed for missing and inconsistent data by a senior researcher. If necessary, clarifying questions or lines of further in- quiry were noted on the returned survey. Special scrutiny was given to contract values and schedule dates to ensure the accuracy of per- formance data. This screening process ensured that follow-up calls with respondents, which were conducted to verify the information for every project, were consistent and professional. Data verifica- tion was conducted with phone calls and e-mails made by research assistants at Penn State and the University of Colorado Boulder. Each assistant was familiarized with the objectives of the research and made aware of the terminology used in the survey. Assistants were trained in verification procedures by observing and participat- ing in several follow-up calls made by a senior researcher. In cases where the survey respondent was not the project owner, the owner was contacted to obtain quality measures and verify the accuracy of cost and schedule data.
A database was developed to enter and store verified survey re- sponses. Data screening was performed to prepare the sample data for analysis. Projects with over 30% missing data across the study variables were removed. Projects that were outside the scope of this research were also removed. These included renovation projects, international projects, civil and highway work, incomplete projects, facilities less than 465 m2 (5,000 gross ft2), and projects achieving final completion prior to 2008. Additionally, several projects that could not be verified with the owner, because of nonresponse from the point of contact, were removed prior to analysis.
Projects with more than one respondent (e.g., an owner and con- tractor submitting a survey for the same project) were combined into a single case. A multistep procedure was followed for resolv- ing discrepancies among respondents on these projects. For quan- titative data, including contract values and schedule dates, the first action was a follow-up discussion with each respondent to verify the data source or confirm that the question was fully understood. If a follow-up was not possible or unsuccessful in resolving the conflict, then the more precise data were used. For example, a sub- stantial completion date given by the contractor as April 2011, was replaced by a more precise date of April 11, 2011, provided by the owner. For qualitative rating data that were collected on ordinal scales, the mean rating of all project respondents was used, with the exception of ratings related to quality. All quality ratings reflected only the opinions of the project owner. For other qualitative data, the study definitions were used to reclassify conflicting responses. For example, general contractors frequently reported that the MEP and structural trades were colocated on-site during construction. During follow-up discussions, the authors discovered that these re- spondents understood colocation to mean having offices on-site, but not necessarily sharing the same office. These responses did not align with the study’s definition of colocation, so they were reclassified prior to analysis.
Analysis of Data
To test the hypotheses, this research considered two types of mod- els: a measurement model and a structural model. The measurement model assessed how well the study’s variables reflected the latent factors they were intended to measure. This assessment was made using a confirmatory factor analysis (CFA). The second type of model, the structural model, was concerned with the relationships between latent factors and was calculated with a series of simulta- neous regression equations. Structural modeling allows for the test- ing of direct and indirect effects, unlike standard regression models,
© ASCE 04016088-3 J. Constr. Eng. Manage.
J. Constr. Eng. Manage., 04016088
D ow
nl oa
de d
fr om
a sc
el ib
ra ry
.o rg
b y
C ar
ne gi
e M
el on
U ni
ve rs
it y
on 1
0/ 04
/1 6.
C op
yr ig
ht A
S C
E . F
or p
er so
na l
us e
on ly
; al
l ri
gh ts
r es
er ve
d.
which are limited to estimating only direct effects. The ability to detect indirect effects in structural models allows us to make substantive statements about the underlying relationships between variables.
Since this study included a combination of ordered categorical and continuous data, the assumption of multivariate normality needed for maximum likelihood (ML) or generalized least-square (GLS) model estimation methods could not be satisfied. Therefore, the model analyses used a mean and variance–adjusted, weighted– least square (WLSMV) regression method. WLSMV estimation incorporates extra nonnegative constants, or weights, for each parameter during the fitting process. Of the estimators suitable for analyzing nonnormal categorical data, WLSMV was found to perform the best in SEM applications (Brown 2006). The weights are derived from the diagonals of the covariance matrix of param- eter estimates. In other words, measurements that are more precise (i.e., having a lower variance) are given more weight, and those with less precision are given less weight. The result is a robust re- gression methodology that provides unbiased parameter estimates for ordered categorical and nonnormal data (Muthén and Kaplan 1985). Missing data were excluded on a per-variable basis using pairwise deletion rules.
Measures
This research established three latent constructs that are hypoth- esized to have an effect on project performance: (1) project delivery methods, (2) team integration, and (3) group cohesion. To select measurement variables for each of these latent constructs, the au- thors investigated literature and vetted potential variables through the study’s advisory board. Where possible, the authors relied on existing measurement variable scales that had been used in prior
project delivery research. The latent constructs and their measure- ment variables are listed in Table 1.
Project Delivery Methods
The classifications of project delivery methods have become blurred somewhat since Konchar and Sanvido’s (1998) study. Over the years, owners have mixed and matched characteristics from DBB, CMR, and DB to suit their project-specific needs. These variations manifest as hybrid contract arrangements, such the U.S. Army Corps of Engineers use of integrated-DBB and more recently the IPD-ish philosophy (El Asmar et al. 2013), neither of which can be clearly classified into a single project delivery method. This study recognizes that traditional classifica- tions of project delivery methods now have limited meaning and clarity in research.
Therefore, the authors used a latent class analysis to identify subsets of project delivery methods. These subsets, or classes, re- present frequently occurring patterns in characteristics associated with the delivery process—specifically, the timing of involvement of contractors, the contract terms, and the procurement practices found within the response data. A complete listing of these char- acteristics, PD1 through PD8, can be found in Table 1. The char- acteristics themselves where translated from the raw questionnaire data into a binary yes or no response to facilitate interpretation. For example, respondents were asked to indicate when, as a percentage of design completion, the builder was contracted. A threshold, such as before or after schematic design, was then used to make the translation. The result of this latent class analysis is the identifica- tion of a categorical latent variable, where each category corre- sponds to a unique set of project delivery method decisions. Each class is then reflected in the presence or absence of the char- acteristics PD1 through PD8. Explained by Franz and Leicht (2016)
Table 1. Measurement Variables Used for the Latent Factors
Latent factor
Measurement variables
Name Description
Project delivery method PD1 Did the owner hold a single contract for design and construction? Yes/No PD2 Was the builder hired during schematic design (SD) or earlier? Yes/No PD3 Were the MEP or structural trades hired during SD or earlier? Yes/No PD4 Was the builder prequalified during procurement? Yes/No PD5 Were the MEP or structural trades prequalified during procurement? Yes/No PD6 Was the builder selected based on cost of work? Yes/No PD7 Were the MEP or structural trades selected based on cost of work? Yes/No PD8 Did the builder have an open-book contract? Yes/No
Team integration I1 Number of BIM uses: 1–5 I2 Proportion of team participating in BIM planning: 0–1 I3 Proportion of team participating in design charrettes: 0–1 I4 Proportion of team participating in goal-setting: 0–1 I5 Proportion of team participating in colocation during construction: 0–1 I6 Off-site prefabrication: Entirely built on-site/entirely built off-site: 1–6
Group cohesion C1 Timeliness of communication: Never on time/always on time C2 Commitment to project goals: Very weakly/very strongly C3 Team chemistry: Poor/excellent C4 Frequency of compromise: Never/frequently C5 Formality of communication: Informal/formal
Turnover experience Q1 Difficulty of startup: 1–6 Q2 Magnitude of callbacks: 1–6 Q3 Operation and maintenance costs: 1–6
Facility quality Q4 Satisfaction with structure and envelope: 1–6 Q5 Satisfaction with interior finishes: 1–6 Q6 Satisfaction with environmental systems: 1–6
© ASCE 04016088-4 J. Constr. Eng. Manage.
J. Constr. Eng. Manage., 04016088
D ow
nl oa
de d
fr om
a sc
el ib
ra ry
.o rg
b y
C ar
ne gi
e M
el on
U ni
ve rs
it y
on 1
0/ 04
/1 6.
C op
yr ig
ht A
S C
E . F
or p
er so
na l
us e
on ly
; al
l ri
gh ts
r es
er ve
d.
and Molenaar et al. (2014), this classification system provides a more data-driven description of how owners are currently deliver- ing their projects. These classes are not intended to represent all variations of project delivery methods, but rather the prevailing pat- terns in characteristics found in our data set. The five classes of project delivery methods derived from the latent class analysis were • Class I: This class was the most “traditional” and directly ana-
logous to a hard bid, design-bid-build approach. The owner held separate contracts for design and construction. The primary builder (CM or general contractor) and specialty subcontractors, specifically the mechanical, electrical, and plumbing (MEP) and structural trades, were hired after design completion and were not prequalified prior to bidding. Construction contracts were issued as lump sums and the award decisions were based exclu- sively on the lowest priced bid. This class was found on 9% of the projects in the sample data set and was the smallest of the five classes.
• Class II: Strong similarities to the delivery described in Class I, with one minor difference: projects in this class required a prequalification step prior to soliciting bids from the primary builder and specialty trade contractors. This class was found on 19% of the sample projects.
• Class III: This class most closely resembles a CMR delivery, where the primary builder is hired during schematic design (SD) or earlier. Specialty trade contractors, however, were not typi- cally contracted until the late stages of design and contracts were awarded to the lowest proposal or bid. Because of the earlier timing of involvement, the primary builder was commonly se- lected on their qualifications, and issued a cost reimbursable contract, most often with a guaranteed maximum price. This class of project delivery was found on approximately one in four projects in the sample (26%).
• Class IV: This class of delivery was comparable to a lump-sum, design-build approach. The owner held a single responsibility contract for design and construction, which was awarded during SD or earlier. A lump-sum contract was awarded to the design- build team that provided the best value of price, qualifications, and experience. This class was found on 27% of sample projects and was the largest of the five classes.
• Class V: Similar to the delivery described in Class IV, with two notable exceptions. First, the owner did not consider the price of work when awarding the contract. The selection was made exclusively on the qualifications and experience of the team. Secondly, the awarded contract had cost reimbursable terms rather than lump sum. Most implementations of integrated pro- ject delivery (IPD) would fall within this class. This class was found on 18% of the projects in the sample data set.
Team Integration
Baiden et al. (2006) describe integrated, interfirm AEC teams as having a singular focus. These teams promote open access to in- formation, give each team member an equal opportunity to con- tribute, and are often colocated together in a common office. Six measures, I1 through I6, were used to evaluate the degree of team integration in project teams for this study. The measures were de- veloped as simple but practical reflections of many behaviors found by Baiden et al. (2006) in highly integrated teams. The measure- ment variables included the following.
The number of BIM uses (I1) is the sum of modeling uses on the project from a predefined list that included architectural design, en- gineered system design, MEP coordination and clash detection, 4D scheduling, and facility management. Since BIM applications often involve information exchanges among multiple team members,
more uses were believed to be reflective of more information sharing across organizations.
Participation in BIM planning (I2) is the proportion of the project team that was involved in developing a BIM execution plan. Execution planning is the process of documenting an implementa- tion method for incorporating BIM into the design, construction, and operation phases of a facility (CIC 2010). The proportion was calculated according to Eq. (1), where the denominator represents the five core project team members: owner, designer, builder, MEP trade contractors, and structural trade contractors. Respondents indicated which of these organizations had at least one represen- tative participating in the BIM planning process. If no BIM exe- cution planning document was generated, the proportion was listed as zero. In a similar manner, proportions for participation in de- sign charrettes (I3), participation in joint goal-setting (I4), and participation in colocation (I5) were calculated for each project. Colocation was defined as two or more team members sharing a common office or workspace. Team members who were on-site at the same time but housed in separate offices were not considered colocated. This measure does not consider the duration of the colocation.
Participation proportion ¼ Number of participating organizations 5
ð1Þ
Off-site prefabrication (I6) is the perceived extent to which building systems were fabricated or modularized off-site and as- sembled on-site. This was evaluated on a semantic differential scale with extremes of entirely built on-site and entirely built off-site. A larger quantity of prefabricated work was thought to require a higher level of coordination among trade contractors that is more likely to occur under an integrated team.
Group Cohesion
The development of strong group cohesion is often viewed as a tipping point, where newly formed work groups finally make the transition to an effective team. Indicative of a shared culture, co- hesion has historically been considered the most important variable in the study of working groups (Carron and Brawley 2000). In organizational research, cohesion is often operationalized with measures of interpersonal attraction, group pride, and task commit- ment. Evidence of cohesion may also be reflected in measures of communication and agreement among members. Five measures, C1 through C5, were used to evaluate the development of group cohesion. All measures were collected on a six-point semantic dif- ferential scale, allowing respondents to indicate both the direction- ality and intensity of their attitudes. The measurement variables included the following.
Timeliness of communication (C1) is the perception that information provided by other team members was received when needed. The scale ranged from never on time to always on time. This measure arises from the concept of information latency, or the lag between when a team member requests information and re- ceives a useful response. Lower latency is associated with higher team satisfaction and reduced schedule durations on projects using an integrated design process (Chachere et al. 2009).
Commitment to project goals (C2) is the perceived extent to which all team members were committed to the same project goals, evaluated on a scale between very weakly and very strongly. Goal commitment is directly analogous to task commitment as a measure of group cohesion (Zaccaro and McCoy 1988).
© ASCE 04016088-5 J. Constr. Eng. Manage.
J. Constr. Eng. Manage., 04016088
D ow
nl oa
de d
fr om
a sc
el ib
ra ry
.o rg
b y
C ar
ne gi
e M
el on
U ni
ve rs
it y
on 1
0/ 04
/1 6.
C op
yr ig
ht A
S C
E . F
or p
er so
na l
us e
on ly
; al
l ri
gh ts
r es
er ve
d.
Team chemistry (C3) is the perception of compatibility among team members, arising from differences in personalities and past and present relationships. This measure originated in prior proj- ect delivery method research (Konchar 1997), and the scale of team chemistry ranged from extremes of poor to excellent. Concep- tually, team chemistry closely aligns with intrapersonal attraction (Festinger et al. 1950) and group pride (Mullen and Copper 1994) in team development literature.
Frequency of compromise (C4) is the perceived prevalence of compromises being made within the project team, evaluated on a scale between never and frequently. To compromise, the project team must collaborate to find a mutually acceptable solution that satisfies all parties. This measure is new to construction research, although the concept of compromise and the balance between co- operation and competition are studied in the field of game theory (Son and Rojas 2011).
Formality of communication (C5) is the perceived extent to which the project team engages in brief, impromptu interaction or more structured, prescribed channels. The scale ranged from informal to formal. Informal communication is viewed as an important mechanism for team members to exchange information interactively and without being scheduled (Kraut et al. 1990). Con- versely, formal communication is typically one-way and follows a preset agenda with a planned list of participants. Both forms of communication have a role on construction projects, but informal interactions are believed to be more supportive of collaborative team relationships.
Project Performance
Every owner views project performance relative to their own project goals. An owner on a new data center project with aggres- sive time-to-market expectations may place greater emphasis on delivery speed, regardless of the extra cost needed for overtime or tenant revisions made near project completion. Therefore, there is no single measure of project performance that is applicable for all building owners. Eight variables were selected to represent cost, schedule, and quality outcomes in this research—six metrics (cost growth, unit cost, intensity, schedule growth, delivery speed, and construction speed) and two latent variables (turnover expe- rience and system quality). With the exception of the two latent variables, each of the metrics were calculated using the equations from Konchar and Sanvido’s (1998) study on project delivery methods.
Costs were collected from the owner’s perspective and were representative of the contractual commitments made for design and construction services. All costs were reported in U.S. dollars and did not include the value of land; permit fees; or furniture, fixtures, and equipment. Cost data were used for three performance metrics: project cost growth, unit cost, and intensity.
Schedule information was collected by requesting both planned and actual dates for the start and finish of several key milestones. Project durations were calculated from the planned and actual dates on each project. These durations were expressed in calendar days and informed three measures of schedule performance: schedule growth, delivery speed, and construction speed.
Quality was measured by asking the owner to rate, relative to their expectations, their satisfaction with the completed facility. These ratings were used additively by Konchar and Sanvido (1998) to represent the overall quality of the project, but were adapted for use in a structural model for this analysis. Shown in Table 1 as variables Q1 through Q6, the questions were grouped into two latent variables: turnover experience and facility quality. Each rating was collected on a six-point semantic
differential scale with extreme values of low and high. This method of measuring project quality is subjective and does not assess the quality of work directly; instead, it reflects the owner’s satisfaction with a project that either met or failed to meet ex- pectations. The owner’s turnover experience was reflected in the difficulty of facility startup (Q1), number and magnitude of callbacks (Q2), and operation and maintenance costs (Q3). Mea- sures of facility quality included the owner’s satisfaction with the envelope and structure (Q4), interior finishes (Q5), and environ- mental systems (Q6).
Data Demographics
Of the 204 projects in the data set, 62% were publicly funded and 38% were private. Facilities were classified into one of nine types describing their intended purpose, including commercial (10% of sample), lodging (13%), office (20%), correctional (2%), educa- tional (27%), manufacturing (5%), sports and recreation (5%), transportation (1%), and health care (16%). The most represented facility type was educational, at 27% of the sample, and the least represented was transportation at only 1%. Projects were distrib- uted across the continental United States, with the largest percent- ages of projects originating from Pennsylvania, Colorado, and California. The sample was geographically broad and not focused on any single region in the United States.
Projects ranged in size from 465 m2 (5,000 ft2) to over 1 mil- lion, although approximately 60% were less than 18,580 m2
(200,000 ft2). Small projects, less than 4,645 m2 (50,000 ft2), represented 27% of the sample. Medium-sized projects between 4,645 m2 (50,000 ft2) and 27,871 m2 (300,000 ft2) accounted for 54%, and large projects greater than 27,871 m2 (300,000 ft2) claimed the remaining 19%. Since the distribution of project size ranges over several orders of magnitude and is positively skewed, logarithmic transformations of base 10 were performed prior to inclusion in the structural model.
The unit cost of projects ranged from $538 per square meter ($50 per square foot) to over $12,917 per square meter ($1,200 per square foot), with 55% reporting less than $4,306 per square meter ($400 per square foot). All unit costs were adjusted for his- torical cost indices and location of the project. Using the Building Cost Index (BCI) reported monthly by the Engineering News Record (ENR), all project costs were adjusted for material and la- bor price fluctuations between their start of construction and June 2014. Location factors provided by RSMeans estimating guides were used to adjust for differences in regional design and construc- tion costs (Waier 2013). For cities that were not directly listed in RSMeans, the location factor for the nearest city with similar economic characteristics was selected.
Results
All model analyses were performed with MPlus statistical soft- ware. For establishing the suitability of both measurement and structural models, two goodness-of-fit indices were used: the root-mean square error of approximation (RMSEA) and compar- ative fit index (CFI). RMSEA measures the error of approxima- tion between the actual data and specified model, relative to the sample size. Well-fitting models have RMSEA less than 0.06 (Hu and Bentler 1995). The CFI compares the specified model against a baseline model that assumes no correlation among variables. The resulting index is a value between 0 and 1, with values at 0.95 and above indicating a good fit (Hu and Bentler 1999).
© ASCE 04016088-6 J. Constr. Eng. Manage.
J. Constr. Eng. Manage., 04016088
D ow
nl oa
de d
fr om
a sc
el ib
ra ry
.o rg
b y
C ar
ne gi
e M
el on
U ni
ve rs
it y
on 1
0/ 04
/1 6.
C op
yr ig
ht A
S C
E . F
or p
er so
na l
us e
on ly
; al
l ri
gh ts
r es
er ve
d.
Measurement Model
A confirmatory factor analysis (CFA) was performed to verify the measurement models for each latent factor. Based on multiple iter- ations and through the elimination of three measurement variables, an acceptable goodness of fit was obtained for the latent factors of team integration and group cohesion. The three eliminated var- iables were formality of communication (C2), frequency of compro- mise (C5), and level of off-site prefabrication (I6). Each of these variables showed low correlations with their latent factors in the CFA. Illustrated in Fig. 2, with standardized coefficients, the final measurement model of group cohesion and team integration is an excellent fit to the data (CFI ¼ 1.0, RMSEA ¼ 0.00). The large standardized regression coefficients on the measurement variables in the group cohesion factor are evidence that team chemistry (0.80), timeliness of communication (0.76), and goal commit- ment (0.83) are all reflective of the same underlying latent factor. Similarly, the CFA produced moderate to large coefficients for variables measuring team integration, including the number of BIM uses (0.72), participation in BIM planning (0.61), goal setting (0.57), design charrettes (0.50), and colocation (0.43). As predicted in the conceptual model, the latent factors of group cohesion and team integration are positively correlated.
Structural Model
All performance metrics were treated as separate dependent vari- ables in the structural model and regressed on the latent constructs of project delivery method, team integration, and group cohesion. The owner type and facility size were included in the model to con- trol for funding source and project scale, respectively. In this man- ner, the significant effects of the latent constructs on performance metrics were separated from the effects of other explanatory var- iables. A p-value of 0.05 was selected as the cutoff for determin- ing statistical significance. The full structural model is presented in Fig. 3, with standardized path estimates. For simplicity, only significant paths are shown in the diagram (p < 0.05) and the classes of project delivery method are represented by a single categorical latent variable. For performance outcomes that varied
by project delivery method, path estimates and significance tests are provided in the table below the model diagram.
Cost Performance
Cost performance was measured using three metrics: project cost growth, unit cost, and intensity. Group cohesion was the only sig- nificant predictor of project cost growth (p ¼ 0.00), when control- ling for project delivery method, team integration, owner type, and facility size, although the overall variation explained was low (R2 ¼ 0.11). In the context of this model, there was no significant difference in mean cost growth across the discrete classes of project delivery methods (0.36 ≤ p ≤ 0.93) and no significant linear rela- tionship between cost growth and team integration (p ¼ 0.79), owner type (p ¼ 0.11), or facility size (p ¼ 0.41).
Both project delivery method and team integration were sig- nificant predictors of unit cost (p ¼ 0.01), when controlling for group cohesion, owner type, and facility size. These relationships explained a total of 26% of the variation in unit cost. Class IV was the only project delivery method that showed signifi- cantly lower unit costs when compared against Class I delivery (p ¼ 0.01), after accounting for differences in group cohesion, team integration, owner type, and gross square footage of the facility. Class V had a lower unit cost than Class I delivery, at a borderline significant level (p ¼ 0.06). Having a public owner resulted in higher unit cost (p ¼ 0.00) and increasing the facility size reduced the unit cost by a small amount (p ¼ 0.01). There was no significant linear relationship between group cohesion and unit cost (p ¼ 0.27). Because of sample size limitations, this model does not control for facility type. Facility type (e.g., com- mercial, lodging, office, etc.) is believed to be a primary factor in determining unit cost, so these results should be interpreted with caution.
Team integration was the only significant predictor of intensity (p ¼ 0.00), when controlling for project delivery method, group cohesion, owner type, and facility size. An increase in team inte- gration saw a corresponding increase in intensity (p ¼ 0.00). Increasing the facility size reduced the overall intensity (p ¼ 0.00) and public owners were associated with more intense projects (p ¼ 0.03), explaining a moderate amount of variation (R2 ¼ 0.37). There was no significant difference in mean intensity across the classes of project delivery methods (0.27 ≤ p ≤ 0.86) and no significant linear relationship between group cohesion and intensity (p ¼ 0.16).
Schedule Performance
Schedule performance was measured with three metrics: project schedule growth, delivery speed, and construction speed. Team integration was the only significant predictor of schedule growth (p ¼ 0.01), with more integrated teams having less schedule growth, when controlling for project delivery method, group cohe- sion, owner type, and facility size, although the overall variation explained was low (R2 ¼ 0.13). Public owners experienced higher schedule growth, at a nearly significant level (p ¼ 0.06). There was no significant difference in mean schedule growth across the classes of project delivery methods (0.07 ≤ p ≤ 0.53), and no sig- nificant linear relationship between group cohesion and schedule growth (p ¼ 0.10) or between facility size and schedule growth (p ¼ 0.37).
Project delivery method, owner type, and facility size were sig- nificant predictors of delivery speed, when controlling for group cohesion and team integration. Within the classes of project deliv- ery methods, there was no significant difference in mean delivery
Fig. 2. Confirmatory factor analysis with standardized estimates
© ASCE 04016088-7 J. Constr. Eng. Manage.
J. Constr. Eng. Manage., 04016088
D ow
nl oa
de d
fr om
a sc
el ib
ra ry
.o rg
b y
C ar
ne gi
e M
el on
U ni
ve rs
it y
on 1
0/ 04
/1 6.
C op
yr ig
ht A
S C
E . F
or p
er so
na l
us e
on ly
; al
l ri
gh ts
r es
er ve
d.
speed between Class II and Class I delivery (p ¼ 0.23). However, when compared against Class I, the mean delivery speeds for Class III (p ¼ 0.01), Class IV (p ¼ 0.00), and Class V (p ¼ 0.00) were significantly faster. Within the context of this model, public
owners were associated with slower delivery speeds (p ¼ 0.01) and larger facilities had faster delivery speeds (p ¼ 0.00). Approxi- mately 86% of the variation in delivery speed was accounted for in this model. There was no significant linear relationship between
Fig. 3. Structural model with standardized estimates
© ASCE 04016088-8 J. Constr. Eng. Manage.
J. Constr. Eng. Manage., 04016088
D ow
nl oa
de d
fr om
a sc
el ib
ra ry
.o rg
b y
C ar
ne gi
e M
el on
U ni
ve rs
it y
on 1
0/ 04
/1 6.
C op
yr ig
ht A
S C
E . F
or p
er so
na l
us e
on ly
; al
l ri
gh ts
r es
er ve
d.
group cohesion and delivery speed (p ¼ 0.97) or team integration and delivery speed (p ¼ 0.53).
Project delivery method, owner type, and facility size were significant predictors of construction speed, when controlling for group cohesion and team integration, explaining a large portion of the variation in construction speed (R2 ¼ 0.84). Within the classes of project delivery methods, there was no significant differ- ence in mean construction speed among Class I, Class II (p ¼ 0.71), and Class III (p ¼ 0.20). Class IV was faster during con- struction (p ¼ 0.00), when compared against Class I. Although not quite significant, the Class V had higher mean construction speed (p ¼ 0.07) than Class I delivery. Within the context of the model, having a public owner resulted in slower construction speed (p ¼ 0.00) and larger facilities had a faster construction speed (p ¼ 0.00). There was no significant relationship between group cohesion and construction speed (p ¼ 0.99) or team integra- tion and construction speed (p ¼ 0.25).
Quality Performance
Quality performance was represented by two latent constructs: turnover experience and overall facility quality. Turnover experi- ence was measured by the owner’s rating on the difficulty of startup, magnitude of callbacks, and operation and maintenance (O&M) costs. The rating scales were oriented such that higher scores of turnover experience signify easier startup, fewer call- backs, and lower O&M costs. Group cohesion and facility size were significant predictors of turnover experience, when control- ling for project delivery method, team integration, and owner type. An increase in group cohesion improved the turnover experience by a significant amount (p ¼ 0.00). For projects with public own- ers, the turnover experience was worse than for private owners (p ¼ 0.00). This model explains 38% of the variation in the owner’s turnover experience. There was no significant difference in turnover experience across the classes of project delivery methods (0.16 ≤ p ≤ 0.99) and no significant linear relationship between team integration and turnover experience (p ¼ 0.54) or facility size and turnover experience (p ¼ 0.30).
Overall facility quality was measured by the owner’s rating of satisfaction with the structure and building envelope, interior fin- ishes, and environmental systems. The rating scales were oriented such that higher levels of facility quality signify greater satisfaction. Group cohesion was the only significant predictor of overall facility quality, when controlling for project delivery method, team integra- tion, owner type, and facility size. An increase in group cohesion was associated with an increase in facility quality (p ¼ 0.00), ex- plaining a moderate amount of variation (R2 ¼ 0.53). Although not quite significant, higher team integration had a positive effect on facility quality (p ¼ 0.06). There was no significant difference in facility quality across the classes of project delivery methods (0.13 ≤ p ≤ 0.78) or with owner type (p ¼ 0.51) and facility size (p ¼ 0.15).
Team Integration
The latent construct for team integration was measured by the number of BIM uses selected by the respondent from a prescribed list and the proportion of the team participating in BIM planning, goal setting, design charrettes, and construction-phase colocation. The rating scales were oriented such that higher scores of team integration were reflective of a higher number of BIM uses and greater proportions of team participation in collaborative interac- tions. Project delivery method, owner type, and facility size were all significant predictors of team integration. Within the classes of
project delivery methods, there was no significant difference in team integration between Class II (p ¼ 0.10) and Class I. In order of increasing means of team integration, Class III, Class IV, and Class V were all significantly different from Class I (p ¼ 0.00). Within the context of this model, having a public owner was asso- ciated with higher team integration (p ¼ 0.05) and larger facilities were delivered by more integrated teams (p ¼ 0.00).
Group Cohesion
The latent construct for group cohesion was measured by ratings of goal commitment, team chemistry, and timeliness of communi- cation. The rating scales were oriented such that higher scores of group cohesiveness were associated with stronger goal commit- ment, better team chemistry, and more frequent on-time communi- cation. The ratings were self-reported by the respondent and reflect their perception of group cohesion after the completion of the project. Project delivery method and team integration were both significant predictors of group cohesion, when controlling for owner type and facility size. Within the classes of project delivery methods, there was no significant difference in mean group cohe- sion among Class I, Class II (p ¼ 0.15), and Class IV (p ¼ 0.11) deliveries. Class V had a higher mean group cohesion score (p ¼ 0.01), when compared against Class I delivery. Although not quite significant, Class III was also associated with a better group cohe- sion (p ¼ 0.06) than Class I delivery. An increase in team integra- tion resulted in a small increase in group cohesion (p ¼ 0.03). This model explained 20% of the variation in the group cohesion on projects in the sample data set. There was no significant linear relationship between owner type and group cohesion (p ¼ 0.59) or facility size and group cohesion (p ¼ 0.42).
Discussion and Applications
Review of the structural model finds support for each of the hypoth- eses. The direct paths to project performance, specifically to project schedule growth, intensity, and group cohesion, from team integra- tion (H1) were positive and significant. Greater group cohesion was directly associated with reduced project cost growth, as well as improved facility quality and turnover experience (H2). However, there was limited support for the hypothesis that project delivery methods directly influence performance (H3), although certain project delivery methods were found to correspond with higher team integration and group cohesion (H3), thereby influencing performance indirectly through these variables. Project delivery methods that involved the builder early in the design process corresponded to higher team integration (H3), while those that in- cluded open-book contract terms and qualification-based selection had a positive and significant direct influence on group cohesion (H3). There was evidence that project delivery methods directly relate to project delivery speed and construction speed; however, there was no support for hypothesized direct-path relationships be- tween project delivery methods and project cost growth or schedule growth (H3).
Prior to this study, the body of literature on the performance of project delivery methods examined only the direct effect of the delivery method itself on various performance metrics. Across multiple studies, DB and CMR are consistently shown to be more effective in controlling cost and schedule growth than traditional DBB (Konchar and Sanvido 1998; Hale et al. 2009; Korkmaz et al. 2010). Demonstrating this relationship, through empirical data col- lection and analysis, was a necessary first step in advancing project delivery research. As a result of these studies, many public agencies revised their procurement policies to allow alternative project
© ASCE 04016088-9 J. Constr. Eng. Manage.
J. Constr. Eng. Manage., 04016088
D ow
nl oa
de d
fr om
a sc
el ib
ra ry
.o rg
b y
C ar
ne gi
e M
el on
U ni
ve rs
it y
on 1
0/ 04
/1 6.
C op
yr ig
ht A
S C
E . F
or p
er so
na l
us e
on ly
; al
l ri
gh ts
r es
er ve
d.
delivery methods. As of 2014, design-build delivery was estimated to hold a 49% market share of nonresidential construction projects in the U.S. (Vashani et al. 2016) and IPD is currently getting wide- spread publicity for its success in the health care sector. However, fundamental questions remain unanswered in literature: Why do some project delivery methods perform better than others? How does the project delivery method affect performance? Answers to these questions will enable more thoughtful evaluation and selection of delivery methods to meet an owner’s project-specific objectives.
This study examined the indirect effect that project delivery methods have on performance, through differences in their level of team integration and group cohesion. The key finding of this study was that project delivery methods do not affect most perfor- mance metrics directly. With the exception of metrics for project delivery speed (R2 ¼ 0.87) and construction speed (R2 ¼ 0.84), the direct effect of the project delivery method on cost, schedule, and quality outcomes was insignificant. Instead, the project deliv- ery method had a significant influence on both the level of team integration (Classes III, IV, and V) and group cohesion (Classes III and V) measured across projects in the data set. This influence was then transmitted indirectly from team integration to metrics of project schedule growth (R2 ¼ 0.13) and intensity (R2 ¼ 0.37), and from group cohesion to project cost growth (R2 ¼ 0.11), fa- cility quality (R2 ¼ 0.53), and the owner’s turnover experience (R2 ¼ 0.38).
A closer look at the commonalities in project delivery methods with significant indirect effects can provide insight. With respect to team integration, Classes III, IV, and V hire the builder, and often the MEP and structural trade contractors, during schematic design or earlier. This suggests that timing of involvement of team mem- bers is a key determinant for team integration. With respect to group cohesion, Classes III and V both use open-book contract terms and primarily use qualifications in the selection process. This is strong evidence that cost transparency and a more relational selection process positively influence the development of project teams. These findings do not refute prior studies that identify a direct link between project delivery methods and performance. Rather, they provide empirical evidence that the concepts of team integration and cohesion address the how and why questions under- lying the performance of project delivery methods.
This study found that project delivery methods explain some, but not all, of the variation in the latent variables of team integration and group cohesion. The combined effect of facility size, owner type, and project delivery method explained 39% of the variation in team integration and 20% of the variation in group cohesion. In other words, there is a statistically significant benefit to team integration and group cohesion that comes from aligning design and construction services with Class III, IV, and V project delivery methods. However, the choice of project delivery method is not the sole determiner of those factors. Construction project teams are complex and comprised of dozens of individuals with their own personalities, backgrounds, and values. Additionally, the organiza- tions to which those individuals belong have their own policies and cultures regulating interactions with other team members (Henisz et al. 2012). The analysis was blind to whether an effective project manager was staffed to a poorly planned project or a careless project manager was assigned to an established, proactive project team. Construction teams are frequently fluid, yet there may be one or two exceptional individuals who drive the jobsite culture and encourage teamwork within an otherwise restrictive structure. Therefore, the project delivery method can be said to provide the potential for more or less integrated teams, but in practice, results
will ultimately be influenced by the individuals and organizations that comprise the project team.
Limitations of the Research
With such a large scope of study, this research acknowledges a few important limitations. First, other external factors may also influ- ence team integration and group cohesion measured on construc- tion projects. Several of those external factors were not practical to collect for the scope of this study. For example, the structural model explained roughly 40% of the variation in team integration. The remaining unexplained variation may be attributed to the owner’s corporate policies that were independent of their project delivery method. An example policy could be an owner requiring BIM use on all projects. If these policies could be reliably col- lected and categorized, it could help to explain the additional variation in the team integration construct. Data on these factors was not obtained, given the format and scope of the data collection process.
Another limitation was not controlling for the facility type in the structural model. Because of the complexity of the structural model specification, the moderate sample size of 204 projects did not al- low for comparisons by facility type. Therefore, the findings of this research reflect relationships found across all sectors of the building construction industry, with the understanding that specific paths may be stronger or weaker depending on the facility type and complexity.
Lastly, a nonresponse study was not conducted. The availability of information on the Internet, and the speed by which it spreads via social media and news outlets, has made building owners more guarded with their project information. This research experienced a response rate of approximately 4% for questionnaires distributed both by e-mail and postal mail. The danger of a low response rate is potentially missing certain subgroups of the population that de- clined to participate. Some reasons for nonresponse were identified by this research during the data verification process, including (1) organizational policies that prohibit sharing of cost information to outside parties, (2) no recent projects within the scope of the study, and (3) too little time to dedicate to the detailed question- naire. While a response rate of 4% is typically considered low for survey research, this rate is consistent with other large-scale project delivery studies. For example, the most directly comparable study performed by Konchar and Sanvido (1998) reported a 5.1% re- sponse rate. However, without a formal nonresponse study, care should be exercised when generalizing the findings of this research.
Conclusions
The primary goal of this research was to model the concept of integration, specifically the aspects of team integration and group cohesion, in the delivery of successful building construction proj- ects. To that end, this research employed several methodologies to represent complex variables and model their relationships, in- cluding a survey questionnaire, latent class analysis, confirmatory factor analysis, and structural equation modeling. As a result of this research, the authors demonstrate that team integration and group cohesion have explanatory value in bridging the causal gap be- tween project delivery methods and project performance. In other words, certain project delivery methods yielded more integrated and cohesive teams, which in turn led to better cost, schedule, and quality outcomes.
Of the five classes of project delivery methods identified in this research, several characteristics were commonly found in more
© ASCE 04016088-10 J. Constr. Eng. Manage.
J. Constr. Eng. Manage., 04016088
D ow
nl oa
de d
fr om
a sc
el ib
ra ry
.o rg
b y
C ar
ne gi
e M
el on
U ni
ve rs
it y
on 1
0/ 04
/1 6.
C op
yr ig
ht A
S C
E . F
or p
er so
na l
us e
on ly
; al
l ri
gh ts
r es
er ve
d.
integrated and cohesive teams. Specifically, deliveries that involved the builder and specialty trades during schematic design or earlier (Classes III, IV, and V) achieved higher levels of team integration. Deliveries that selected the builder based on qualifications only and issued an open-book, cost plus or guaranteed maximum price (GMP) contract (Classes III and V) were associated with more co- hesive teams. Conversely, delivery methods with late involvement of the builder and specialty trades, strictly price-based selection and closed-book, lump-sum contracts (Classes I and II) produced both the least integrated and least cohesive teams. Additionally, when allowing for indirect effects through team integration and group cohesion, project delivery methods only had a significant direct effect on project delivery speed and construction speed. Deliver- ies that involved the builder during schematic design or earlier (Classes III, IV, and V) saw a large percent increase in average project delivery speed. However, only deliveries that also involved specialty trades during schematic design or earlier (Class IVand V) saw a corresponding increase in construction speed.
This study represented the concepts of team integration and group cohesion as latent variables, and both were found to influ- ence distinct project outcomes. Integrated teams involved all tiers of project participants—from designers to specialty trades—in design charrettes, goal setting, BIM applications and planning, and colocation during construction. On average, as team integration in- creased, project schedule growth was reduced and project intensity was increased. Cohesive teams were generally more developed, reporting higher chemistry, commitment to project goals, and time- liness of communication. Improvements in group cohesion resulted in reduced project cost growth, as well as higher owner ratings for building system quality and the turnover experience. This study corroborates the literature on the positive effects of teamwork on project performance (Dietrich et al. 2010; Baiden and Price 2011). Future studies may expand this methodology to determine the ef- fect of team integration and group cohesion on both sustainability and safety performance.
Prior empirical research on project delivery methods was highly influential on policymakers and led to increased adoption of de- sign-build in the United States. However, that research focused almost exclusively on the delivery methods themselves, when they are not the sole contributor to project success. This paper demon- strates that project delivery methods need to be considered along- side other factors, specifically team integration and group cohesion, to better understand their relationship with project performance. In this way, the concept of integration can begin to be decoupled from project delivery. Instead of expecting the delivery method to automatically lead to success, project participants can take a more active role in developing as a team. This means viewing the con- struction project as a long-term relationship, where careful consid- eration is given to the addition of new team members and efforts are made to align interfirm efforts, regardless of the delivery method being used.
Acknowledgments
The authors extend appreciation to the primary research sponsor, the Charles Pankow Foundation. They also acknowledge a contri- bution from the Construction Industry Institute (CII), thoughtful feedback provided by the industry advisory board, and the efforts of research assistants at Penn State and University of Colorado Boulder: Behzad Esmaelli, Lars Anderson, Kayleigh Arendt, Bryan Doyle, Alexander Van Melle, Rachel Sommer, Shelby White, and Jared Zoller.
References
Baiden, B., and Price, A. (2011). “The effect of integration on project delivery team effectiveness.” Int. J. Project Manage., 29(2), 129–136.
Baiden, B., Price, A., and Dainty, A. (2006). “The extent of team inte- gration within construction projects.” Int. J. Project Manage., 24(1), 13–23.
Bogus, S., Shane, J., and Molenaar, K. (2010). “Contract payment provi- sions and project performance: An analysis of municipal water and wastewater facilities.” J. Publ. Works Manage. Policy, 15(1), 20–31.
Bromley, S., Worthington, J., and Robinson, C. (2003). The impact of integrated teams on the design process, Construction Productivity Network, London.
Brown, T. (2006). Confirmatory factor analysis for applied research, Guilford Press, New York.
Carron, A., and Brawley, L. (2000). “Cohesion conceptual and measure- ment issues.” Small Group Res., 31(1), 89–106.
Chachere, J., Kunz, J., and Levitt, R. (2009). “The role of reduced latency in integrated concurrent engineering.” ⟨http://cife.stanford.edu/sites/ default/files/WP116.pdf⟩ (Dec. 7, 2015).
Chan, D., Chan, A., Lam, P., Lam, E., and Wong, J. (2007). “An investi- gation of guaranteed maximum price (GMP) and target cost contracting (TCC) procurement strategies in the Hong Kong construction industry.” J. Financial Manage. Property Constr., 12(3), 139–150.
Chinowsky, P., Diekmann, J., and O’Brien, J. (2010). “Project organiza- tions as social networks.” J. Constr. Manage. Eng., 10.1061/(ASCE) CO.1943-7862.0000161, 452–458.
CIC (Computer Integrated Construction) Research Program. (2010). BIM project execution planning guide—Version 2.0, Pennsylvania State Univ., University Park, PA.
Dainty, A., Briscoe, G., and Millet, S. (2001). “New perspectives on construction supply chain integration.” Supply Chain Manage.: An Int. J., 6(4), 163–173.
Dietrich, P., Eskerod, P., Dalcher, D., and Sandhawalia, B. (2010). “The dynamics of collaboration in multipartner projects.” Project Manage. J., 41(4), 59–78.
El Asmar, M., Hanna, A., and Loh, W. (2013). “Quantifying performance for the integrated project delivery system as compared to established delivery systems.” J. Constr. Eng. Manage., 10.1061/(ASCE)CO.1943 -7862.0000744, 04013012.
El Wardani, M., Messner, J., and Horman, M. (2006). “Comparing pro- curement methods for design-build projects.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-9364(2006)132:3(230), 230–238.
Evbuomwana, N., and Anumbab, C. (1998). “An integrated framework for concurrent life-cycle design and construction.” Adv. Eng. Software, 29(7-9), 587–597.
Festinger, L., Schachter, S., and Back, K. (1950). “The spatial ecology of group formation.” Chapter 4, Social pressure in informal groups, L. Festinger, S. Schachter, and K. Back, eds., Harper, New York.
Franz, B., and Leicht, R. (2016). “An alternative classification of project delivery methods used in the United States building construction industry.” Const. Manage. Econ., 34(3), 160–173.
Hale, D., Shrestha, P., Gibson, G., Jr., and Migliaccio, G. (2009). “Empiri- cal comparison of design/build and design/bid/build project delivery methods.” J. Constr. Eng. Manage., 10.1061/(ASCE)CO.1943-7862 .0000017, 579–587.
Henisz, W., Levitt, R., and Scott, R. (2012). “Toward a unified theory of project governance: Economic, sociological and psychological supports for relational contracting.” Eng. Project Organiz. J., 2(1-2), 37–55.
Hu, L., and Bentler, P. (1995). “Evaluating model fit.” Structural equation modeling: Concepts, issues, and applications, R. H. Hoyle, ed., Sage, London, 76–99.
Hu, L., and Bentler, P. (1999). “Cutoff criteria for fit indexes in covari- ance structure analysis: Conventional criteria versus new alternatives.” Struct. Equ. Model.: A Multi. J., 6(1), 1–55.
Kent, D., and Becerik-Gerber, B. (2010). “Understanding construction industry experience and attitudes toward integrated project delivery.” J. Constr. Eng. Manage., 10.1061/(ASCE)CO.1943-7862.0000188, 815–825.
© ASCE 04016088-11 J. Constr. Eng. Manage.
J. Constr. Eng. Manage., 04016088
D ow
nl oa
de d
fr om
a sc
el ib
ra ry
.o rg
b y
C ar
ne gi
e M
el on
U ni
ve rs
it y
on 1
0/ 04
/1 6.
C op
yr ig
ht A
S C
E . F
or p
er so
na l
us e
on ly
; al
l ri
gh ts
r es
er ve
d.
Konchar, M., and Sanvido, V. (1998). “Comparison of U.S. project delivery systems.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-9364(1998) 124:6(435), 435–444.
Konchar, S. (1997). “A comparison of United States project delivery sys- tems.” Technical Rep. No. 38, Penn State’s Computer Integrated Con- struction (CIC) Research Program, University Park, PA.
Korkmaz, S., Riley, D., and Horman, M. (2010). “Piloting evaluation met- rics for sustainable high-performance project delivery.” J. Constr. Eng. Manage., 10.1061/(ASCE)CO.1943-7862.0000195, 877–885.
Kraut, R., Fish, R., Root, R., and Chalfonte, B. (1990). “Informal com- munication in organizations: Form, function and technology.” Human Reactions to Technology: Claremont Symp. on Applied Social Psychology, Sage, Beverly Hills, CA, 145–199.
Kumaraswamy, M., Ling, F., Rahman, M., and Phng, S. (2005). “Constructing relationally integrated teams.” J. Constr. Eng. Manage., 10.1061/(ASCE)0733-9364(2005)131:10(1076), 1076–1086.
Love, P., Gunasekaran, A., and Li, H. (1998). “Concurrent engineering: A strategy for procuring construction projects.” Int. J. Project Manage., 16(6), 375–383.
Molenaar, K., Messner, J., Leicht, R., Franz, B., and Esmaeili, B. (2014). “Examining the roles of integration in the success of building construc- tion projects.” ⟨http://www.pankowfoundation.org/grants.cfm⟩ (Apr. 11, 2015).
Mollaoglu-Korkmaz, S., Swarup, L., and Riley, D. (2013). “Delivering sustainable, high-performance buildings: Influence of project deliv- ery methods on integration and project outcomes.” J. Manage. Eng., 10.1061/(ASCE)ME.1943-5479.0000114, 71–78.
Moore, D., and Dainty, A. (2001). “Intra-team boundaries as inhibitors of performance improvement in UK design and build projects: A call for change.” Constr. Manage. Econ., 19(6), 559–562.
MPlus version 7.2 [Computer software]. Linda Muthén and Bengt Muthén, Los Angeles.
Mullen, B., and Copper, C. (1994). “The relation between group cohe- siveness and performance: An integration.” Psychol. Bull., 115(2), 210–227.
Muthén, B., and Kaplan, D. (1985). “A comparison of some methodologies for the factor-analysis of non-normal Likert variables.” Br. J. Math. Stat. Psychol., 38(2), 171–189.
Smyth, H., and Pryke, S. (2008). Collaborative relationships in construc- tion: Developing frameworks and networks, Blackwell Publishing, West Sussex, U.K.
Son, J., and Rojas, E. (2011). “Evolution of collaboration in temporary project teams: An agent-based modeling and simulation approach.” J. Constr. Eng. Manage., 10.1061/(ASCE)CO.1943-7862.0000331, 619–628.
Vashani, H., Sullivan, J., and El Asmar, M. (2016). “DB 2020: Analyzing and forecasting design-build market trends.” J. Constr. Eng. Manage., 10.1061/(ASCE)CO.1943-7862.0001113, 04016008.
Waier, P. (2013). “RSMeans building construction cost data 2014.” Reed Construction Data, LLC, Norwell, MA.
Walker, D., and Hampson, K. (2003). Procurement strategies: A relationship-based approach, Blackwell Science, Malden, MA.
Zaccaro, S., and McCoy, M. (1988). “The effects of task and interpersonal cohesiveness on performance of a disjunctive group task.” J. Appl. Soc. Psychol., 18(10), 837–851.
© ASCE 04016088-12 J. Constr. Eng. Manage.
J. Constr. Eng. Manage., 04016088
D ow
nl oa
de d
fr om
a sc
el ib
ra ry
.o rg
b y
C ar
ne gi
e M
el on
U ni
ve rs
it y
on 1
0/ 04
/1 6.
C op
yr ig
ht A
S C
E . F
or p
er so
na l
us e
on ly
; al
l ri
gh ts
r es
er ve
d.