1. Introduction
Prefabricated construction (PC) technology is an innovative construction technology that involves the transfer of on-site construction work from the site to the factory [
1,
2]. PC technology is a resource-saving and environment-friendly construction technology that addresses many important problems, such as construction noise, waste pollution, and resource waste [
3,
4]. Over the past three decades, PC technology has received increasing attention worldwide, and the adoption of PC technology has made considerable progress [
5,
6]. For example, in early 1996, the prefabrication levels in Germany, the Netherlands, Denmark, and Sweden rose by 31%, 40%, 43%, and 80%, respectively [
7]. The size of the UK’s PC industry increased from £2.2 billion in 2004 to £6 billion in 2006 [
8]. In Asia, Singapore is the first country to formulate industry codes and regulatory frameworks for prefabricated construction projects (PCPs) [
9].
The development of PC in developed countries shows that PC is an effective driving force for the growth of the construction industry [
10,
11]. The positive effects of PC are manifested in improving construction quality and safety, reducing labor demand, shortening construction periods, and improving environmental performance [
12,
13]. Although these advantages largely meet the requirements of urbanization, housing needs, and sustainability in developing countries, PC is not widespread and mature, for instance in China [
8]. Compared with most Western countries, China’s PC industry is still in its infancy [
14]. In 2022, China’s total construction area reached 4.3 billion square meters, while the prefabricated construction area was 1.08 billion square meters, accounting for only 25% of China’s total construction area. This ratio is far less than that in other developed countries, such as the United States (90%), France (85%), Sweden (80%), and Denmark (80%). The Chinese government is very concerned about unsustainable practices in construction activities, and expects to heavily promote the implementation of PC through policies and regulations [
15].
PC is a cross-organizational innovation [
16,
17]. Implementing PC in projects requires multiple industry organizations (e.g., owners, contractors, designers, suppliers) to participate and collaborate [
18,
19]. As different prefabricated construction projects are initiated and implemented, industry organizations will gradually gather into a complex and evolving collaborative relationship network at the industry level based on specific collaboration relationships with different project backgrounds [
20,
21]. This industry-level, evolving collaboration network is an important channel for PC technology’s diffusion [
22,
23]. Investigating the evolution of industry-level collaborative networks for PC technology implementation can reveal how industry organizations interact with each other in PCP implementation practices and how PC technology diffuses among industry organizations through project-based collaborative ties. This provides theoretical guidance for PC industry organizations to develop efficient collaboration strategies, as well as a decision-making reference for government agencies to formulate appropriate industry incentive strategies, thereby contributing to the development of the PC industry.
Studies over the past decades have focused on theoretical or empirical research concerning problems related to PC technology in practice and the specific project network of PC technology implementation. Specifically, these studies focused on: (1) identifying the potential network of risk factors related to PCP stakeholders [
16,
19]; (2) investigating the constraints and interactions between different tasks in PCP implementation [
12,
20]; and (3) exploring the macro structural characteristics of the patent collaboration network in the PC field [
18]. However, the study of the evolution of the industry-level collaboration network for PC technology implementation is lacking. This study applied the stochastic actor-oriented models (SAOM) method and used a longitudinal data set of PCPs in Shanghai, China from 2015–2023, with the following goals: (1) explore how macro structure of the project-based collaboration network for PC technology implementation evolves over time; and (2) determine how corresponding micro effects drive the evolution of the macro network structure. As an exploratory effort of using network dynamics models to investigate the self-organizing mechanisms of the evolution of inter-organizational collaboration networks in the PCP field, this study contributes to the complex adaptive system theory, the growing literature on PCPs, and the existing knowledge body of collaborative networks in the construction domain. The results are expected to help PC industry organizations to formulate collaboration strategies and government agencies to develop incentive strategies for the PC industry.
3. Research Method
3.1. Data Source
This study uses longitudinal data for construction projects using PC technology in Shanghai from 2015 to 2023 to investigate the evolution of the project-based collaboration network for PC technology implementation over time and to assess the proposed hypotheses. Shanghai is China’s largest city, with a construction output value of 139.605 billion dollars in 2023; it is one of the first regions in China to promote the application of PC technology. The Shanghai Housing and Urban-rural Construction Administration Commission issued the “Development Plan of Shanghai Prefabricated Buildings” in 2016, stating that by 2025, PC should be one of the main construction modes in Shanghai to increase environmental benefits.
The PCP data for this study come from the Shanghai Construction Trade Association (SCTA). The data include the project name, project scale, project type, project start year, project location, owner, prime contractor, supplier, and designer. Specifically, PCP data include data for 236 projects involving 316 organizations for the time period from 2015 to 2023. The year 2015 serves as the first observation time point, because the official statistics related to Shanghai PCPs are incomplete before 2015. Incomplete data would result in insufficient and unrepresentative network data, hindering a meaningful network analysis.
Table 2 provides the demographic information of these projects and related organizations.
Table 2 shows that the 236 PCPs differ with respect to project type, project size, and project starting year. A total of 316 organizations are reported to have been part of these projects, including 173 owners, 56 designers, 48 prime contractors, and 39 suppliers. These organizations are used for subsequent network analysis (i.e., descriptive analysis and SAOM analysis). In descriptive analysis, this study uses nine year-long time windows (2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, and 2023) to analyze the changes in the collaboration relationship from the perspective of the network. In SAOM analysis, this study investigates preferential attachment effect, geographic proximity effect, ownership similarity effect, experience similarity effect, and size similarity effect.
Figure 1 shows the specific research process.
Owner choice drives the establishment of collaborative relationships between organizations. The ties between different nodes in the network reflect the collaborative relationship between different organizations. This study uses a one-mode network (i.e., using organizations as network nodes) to describe the inter-organizational collaboration networks. The study data are represented in a one-mode matrix. This matrix is represented by mij, where i and j represent two different organizations. Each year-long window corresponds to a matrix, and each cell mij in the matrix represents the number of collaborations between two organizations. The matrices also describe the direction of collaboration ties between organizations. Specifically, cell mij describes that organization i has a directional tie to organization j; in other words, organization i is the sender of the collaboration tie, and organization j is the receiver of the collaboration tie. The formation of the collaborative relationships between owners, designers, prime contractors, and suppliers depends on how owners choose the other partners. As such, the study matrices only describe directional ties from owners to designers, prime contractors, and suppliers. Therefore, the network structure is asymmetric: there are no two-way ties between network nodes. In addition, the subsequent data processing in SAOM analysis also follows the setting of the one-mode network mode.
3.2. Descriptive Analysis
To begin the analysis, the macro structure of the collaboration network under different time windows is described. This is followed by the SAOM analysis, testing the hypotheses related to the potential mechanism of the network evolution (i.e., quantifying the micro mechanism of network evolution). Based on previous studies on inter-organizational collaboration in the construction field [
21,
23], 9 network indicators are used to describe the characteristics of the collaboration network structure under different time windows (see
Table 3).
3.3. Stochastic Actor-Oriented Models Analysis
SAOM is used to analyze the evolution of collaborative networks. It allows for the simultaneous analysis of different mechanisms driving the evolution of the network, and is considered an effective method for processing longitudinal network data [
45,
58]. As a role-based modeling method, SAOM applies the following assumptions. First, the evolution of the network structure takes the form of a Markov chain, meaning that the network state at time t is determined by the network state at time t − 1, and not the network state at time t − 2 or t − 3. Second, the potential time parameter t is continuous, and the observable change (i.e., wave in SAOM) of the network structure from one time point to the next is completed through a large number of micro-level probabilistic and continuous small steps. In each small step, the network actor can change only one outgoing tie. Third, SAOM analyzes directional ties. Based on these assumptions, a computer is used to simulate the statistical process of the network changes. The first task in the simulation is to determine the actors allowed to make changes in each small step. These actors make choices, which include changing an outgoing tie or not making changes to the outgoing ties. The probability of making choices is modeled as an objective function, representing the possibility that the actor changes outgoing ties. The objective function in SAOM is modeled as a set of linear combinations of network structure-based effects and node attribute-based effects.
In this expression, indicates that the value of the objective function of the actor depends on the state of the network; indicates the effect that affects the change of the network connection; and indicates the statistical parameter of each effect. includes network structure-based effects related to state of the network and actor attribute-based effects. An iterative Markov chain Monte Carlo algorithm is used to evaluate the weight of . If , the corresponding effect has no influence on the network dynamics. A value of indicates a greater probability that the network evolves in the direction of a higher corresponding effect. A value of indicates a smaller probability that the network evolves in the direction of a higher corresponding effect.
5. Discussion
This study descriptively analyzes how the macro structure of the collaboration network for PC technology implementation evolves over time, and investigates how related micro mechanisms jointly support the evolution of the network. The results show that the collaboration network becomes denser over time, and continues to show a core–peripheral structure with a small number of super-connected nodes. In terms of the micro mechanism, the experience similarity effect and the size similarity effect are not significant. This differs from the conclusions of Li et al. [
33] and Tang et al. [
23]. The results indicate that the tendency of owners and designers/prime contractors/suppliers with similar organizational size/organizational experience to form new project-based partnerships is not significant. This may be because the PCP experience held by the owners is not an important factor impacting whether prime contractors, designers, and suppliers participate in project bidding. In contrast, these prime contractors, designers, and suppliers may be more concerned about other factors, such as the owner’s reputation and influence, contract value, and the social impact of the project [
22]. Similarity in organization size is not a key factor impacting the owner’s choice of partners. Instead, the owner may focus more on the professional ability, management ability, organizational ability, and operational efficiency of prime contractors, designers, and suppliers [
21]. Collaboration between organizations of different sizes can help integrate complementary resources. Large organizations obtain three key benefits by collaborating with medium and small-sized organizations: (1) accessing partners’ professional areas and technologies; (2) increasing project flexibility; and (3) sharing risks associated with PCPs [
41]. Medium- and small-sized organizations can obtain two key benefits by collaborating with large organizations: (1) mitigating a lack of experience with risk management; and (2) forming long-term collaborative relationships with each other, such as strategic alliances [
46]. These may improve the overall competitiveness of industry organizations and their performance.
The significant preferential attachment effect occurs because owners tend to select prime contractors, designers, and suppliers with prominent network position as partners (i.e., having high influence); this reduces the burden of project risks and encourages smooth PCP implementation. Highly influential partners build confidence for the owners and are easier for the owners to trust. Meanwhile, partners with prominent network positions usually have good professional capabilities, reputation, and contract success, increasing the chance of a smooth project delivery [
61]. The explanation for the significant similarity effect with respect to organizational ownership is that PCPs require more innovative knowledge and technical input than traditional projects. While it may seem counter-intuitive that more homogenous ownership models may increase innovation, it is important to understand the state-owned organization framework in China. State-owned organizations benefit from the support of state funds, are more willing to focus on learning and technological innovation, and are generally more willing to take risks. Previous studies have also found that the technological innovation investment of state-owned enterprises is an important factor in encouraging the implementation of emerging technologies [
21]. Furthermore, state-owned organizations in China are generally willing to take on more social responsibilities than non-state-owned organizations. Specifically, they are willing to develop new technologies (e.g., PC technology) that save resources and reduce the negative impact of construction projects on society and the environment [
55]. Further, Chinese law stipulates that state-owned enterprises cannot become general partners with unlimited liability, while non-state-owned enterprises need to assume unlimited liability. Therefore, there are differences between China’s state-owned enterprises and non-state-owned enterprises with respect to innovation, risk preference, assumed liability, and the assumption of social responsibility. These factors somewhat limit their collaboration.
The result with respect to the significant geographic proximity effect is similar to the conclusion of Tang et al. [
23]. In construction projects, having a closer geographic distance encourages the formation of inter-organizational collaboration relationships. First, geographical proximity can reduce transportation costs (e.g., the closer the geographical distance between contractors and suppliers, the lower the transportation cost of building components). Second, geographical proximity facilitates face-to-face communication between organizations. This helps resolve conflicts between the organizations and focuses the attention of the organization personnel on project tasks. It can also increase each organization’s understanding of the other’s work content and difficulties, building mutual trust between organizations. Third, geographically close organizations generally have more information about each other (e.g., reputation, organizational capabilities, management capabilities). This reduces the cost of information in the partner selection phases, the risk of a poor selection, and the subsequent coordination costs in project implementation. Fourth, geographically close organizations can also facilitate the smooth project implementations by jointly addressing emergencies that arise during PCP implementation. In summary, this study’s findings clarify the evolutionary rules impacting the PCP collaboration network structure, and confirm how related micro mechanisms collectively drive the evolution of the network.
6. Theoretical Implications
This research has the following theoretical implications. First, by using the network perspective to analyze the collaboration relationships with respect to PCP implementation, this research helps deepen understanding of how industry organizations interact with each other across PCPs to form project-based, inter-organizational collaboration networks. This contribution is important because previous research on PCPs mainly focused on analyzing problems related to PCP practices and specific project networks constructed for PC technology implementation. However, PC technology practices have not been studied at the industry level. This study identified an increase in the density of the PCP collaboration network and a continuous core–peripheral structure of the network; this indicates that the distribution of inter-organizational collaboration relationships is continuously uneven. There are super-connected organizations in the network, with many collaboration relationships with other industry organizations. These network characteristics may accelerate the diffusion of PC technology-related knowledge and innovations, but may also lead to continuous unbalanced inter-organizational collaboration relationship. In the long term, this may hinder the diffusion of PC technology among organizations and the industry-level development of inter-organizational collaboration networks for PC technology implementation. This also helps explain why the popularization of PC technology still faces challenges, despite being vigorously promoted by the government.
Second, previous studies of the industry-level inter-organizational collaboration networks in the construction industry have expanded an understanding of the static structure of the collaboration network in specific periods and a regional context [
62,
63]. However, the dynamic evolution process of collaboration network and the micro mechanisms that support the network evolution have remained unclear. In contrast, this research explores using a network dynamics model to quantitatively explore how micro mechanisms drive the evolution of project-based collaboration networks. The study provides evidence that the evolution of the collaboration network relates to the preferential attachment effect, the geographic proximity effect, and the organizational ownership similarity effect. This research helps deepen an understanding of the micro foundation of the evolution of inter-organizational collaboration relationships, and emphasizes the need to consider the collaboration network as a complex adaptive system. The dynamics of this system are closely related to a set of self-organization effects (i.e., similarity effects, attachment effects).
7. Managerial Implications
This research has the following managerial implications. First, in the collaboration network for PC technology implementation, the preference attachment effect is an important mechanism; this effect impacts the network position of organizations and the evolution of collaboration network. Therefore, industry organizations facing technological innovation or market changes, including designers, prime contractors, and suppliers, can exercise first-mover strategies to occupy favorable network positions in the industry, increasing organizational influence and competition force [
23].
Second, the structural characteristics and evolutionary laws of the project-based collaboration network for PC technology implementation are closely related to the diffusion of PC-related knowledge and innovation in the construction industry. Policy-setting agencies need to understand these characteristics and evolutionary laws when designing policy to manage industry organizations and encourage the innovation of PC technology and the diffusion of PC-related knowledge. This research provides evidence that collaborative networks usually develop around super-connected organizations. This leads to an uneven distribution of collaboration relationships between industry organizations. As such, intervention policies could be developed for these star organizations, such as appropriately limiting the number of these organizations participating in PCPs each year. Incentive policies could be developed for other industry organizations, such as reducing taxes on PCPs and providing subsidies for those who create more balanced inter-organizational collaboration networks [
22].
Third, the ownership similarity effect found in this study indicates that the tendency toward homogeneity may cause the separation of organizations with different ownership models (i.e., state-owned organizations, non-state-owned organizations), discouraging industry resource integration. Therefore, policy-setting agencies can take measures to ease this situation and promote collaboration between organizations with different ownership models. Incentive policies that would encourage collaboration between industry organizations with different ownership models could include providing project rewards to diverse teams, increasing the effective diffusion of PC technology-related knowledge and innovation in the industry. The significant geographic proximity effect also has implications for policy-setting agencies; as such, the government could provide subsidies for geographically-diverse projects and establish long-distance communication channels, including PC technology-related forums or training, to promote communication and collaboration.
9. Limitations and Future Work
This study has two limitations to address in future studies. First, the study was conducted within the specific market context of Shanghai, China. This may somewhat limit the applicability of the research results in other market settings. Future research should further expand the research and explore the dynamics of collaboration networks for PC technology implementation in different market contexts. Second, based on data availability, this study focused on the relationship between owners, designers, prime contractors, and suppliers. Future research should cover more organizations, including consultants. This would expand the understanding of the collaboration network of PCPs.