Network analysis of interconnections between theoretical concepts associated with principal–agent theory concerning construction projects

In the field of construction management, principal–agent (P–A) theory has received increased attention in recent years, as it offers a suitable representation of project management in construction. Researchers have contributed relevant studies to the literature. Sha (2019) explored the triangular relationship between the client, contractor and project manager; Lin and Wang (2019) used P–A theory to evaluate dynamic knowledge-sharing in construction project teams; Hosseinian et al. (2020) explored incentive contracts in a model with multiple agents; Cerić (2019) explained the relationship between P–A theory and block-chain to minimise communication risks.

In construction projects, there are many participants or stakeholders, with a wide variety of expertise and interests. Their interactions are undoubtedly subject to information asymmetry, which has implications in P–A theory. George Akerlof, Michael Spence and Joseph Stiglitz shared a Nobel Prize in Economics in 2001 for their work on information asymmetry conducted in the 1970s. Information asymmetry describes a condition in which one party is better informed than another, and the parties do not share the same interests (Jensen, 2000). Information asymmetry can cause communication risks in all project phases.

Information asymmetry can result from the self-interest of the participants or the unwillingness of stakeholders to share information. According to P–A theory, there are three types of information asymmetry: hidden characteristics, hidden information and hidden intention. These generate the following risks: adverse selection, moral hazard and hold-up (Jäger, 2008; Schieg, 2008). Adverse selection may occur if the principal does not know the exact qualifications of the agent before signing a contract. A moral hazard may occur if the principal is unsure that the agent will fully act on his/her behalf after a contract is signed. A hold-up can occur if the agent acts opportunistically after the principal has invested resources in the belief that the agent will act appropriately.

In early studies in the field of project management (Turner and Müller, 2004; Müller and Turner, 2005), the main P–A relationship was between the project owner and the project manager. Research has shown that the behaviour of project managers during the construction phase is especially important for project success, more than that of project owners and contractors (Cerić, 2016). Other participants engaged in construction including contractors, sub-contractors, designers and consultants can also be considered as agents. The most common assumption is that the project owner is the principal and the contractor is the agent. According to P–A theory, principals and agents are guided by self-interest and opportunism. Adverse behaviour is a characteristic of both parties, but the opportunism of agents has been explored in greater depth because it is expected that agents will try to maximise their benefits even at the expense of the principal (Schieg, 2008). As the number of involved parties grows and the number of non-contractual relationships between them increases, the mechanisms of formal control become less capable of suppressing opportunistic behaviour (Cerić, 2016).

This study analyses the incidence of P–A theory in construction management literature with the purpose of assessing new development on the subject. The literature analysis was conducted using the keywords listed by the authors. Keywords analysis was chosen because keywords accurately represent the theme of a paper. Focusing on the leading journals in the field, this study builds on a previous review by Cerić (2013) and investigates the connection between P–A theory and its areas of application with deeper historical roots in construction management. Seven leading journals in the construction management field were searched using ‘principal–agent’ as the keyword. ‘Principal–agent’ is most often followed by ‘theory’, but ‘problem’ and ‘model’ are also associated in the literature. Hereafter, only the word ‘theory’ is considered.

This paper is organised into four sections. The methodology of the literature review is explained in Section 2. In Section 3, the findings are highlighted, mainly through tables and network diagrams. A discussion is presented in Section 4. The conclusions and limitations of the study, and suggestions for future research in this area are presented in Section 5. This review is intended to enrich construction management research and present new perspectives on the interconnections between theoretical concepts associated with P–A theory concerning construction projects.

The presented research was collected in three steps: selection of journals, keyword search and keyword analysis. The journals to be investigated were selected first. There are two sampling methods for publications used in literature reviews. Some researchers choose leading journals to conduct a literature search (Xia et al., 2018). Others use common computer search engines and established literature databases (Sabini et al., 2019). The aim of this study was to investigate the extent to which P–A theory is used in construction management research. A literature search was conducted focusing on the leading journals in the field of construction management. The main keyword for the search was ‘principal–agent’, and this keyword was used in the identification of papers relevant to P–A theory. All associated keywords in the relevant papers were identified and analysed for interconnections, which is the main contribution of this review. Keyword analysis is important because keywords represent the theme of a paper.

This literature review is based on Bröchner and Björk (2008). Through an opinion survey, their research identified seven leading journals in the field of construction management. The seven leading journals in construction management identified by Bröchner and Björk, and their publishers (2008, p. 742), are shown in Table 1. For the purposes of this literature review, each of their online archives was searched using the keyword ‘principal–agent’.

To map the concepts related to P–A theory in construction management research and highlight the key relationships, this study provides a network analysis (NA) of the relevant keywords.

In addition to use in social research to determine relational data, NA is conducted in project planning, complex systems, electrical circuits, transportation systems, communication networks, epidemiology, bioinformatics, hypertext systems, text analysis, bibliometrics, organisational theory, genealogical research and event analysis (Brandes and Erlebach, 2005). In element-level analysis (Brandes and Erlebach, 2005), the main questions are: ‘Which is the most important element?’ and ‘How important is this element?’ In this study, the element is a keyword, meaning a concept connected to P–A theory in the construction management literature.

Gephi software was used to conduct the NA. Gephi is an open-source software for graph and NA that uses a 3-D render engine to display networks in real time for faster statistical analysis (Bastian et al., 2009). In NA, relationship data are visually represented and statistically evaluated. A visual representation is created; all identified keywords are depicted as nodes, and the interconnections between them are shown as edges. In this case, weights are assigned to each edge to describe the importance of the specific relationship between keywords. In NA, edge weights can be represented as a function ω : E → R that assigns each edge e ∈ E a weight ω (e) (Brandes and Erlebach, 2005). Depending on the context, edge weights can describe properties such as cost, capacity and strength of interaction, or in this case, the number of times that a specific connection appears in the analysed papers. All input data are available in the Appendix. This network is undirected because the interconnection between two keywords is a two-way relationship. Network visualisation enables the mapping of the keywords; statistical analysis enables the mapping of key relationships between keywords. Only some of the network properties were calculated, as the main objective was the network visualisation. These network properties include the following:

Density: an indicator of network connectivity describing the portion of potential connections in a network that are actual connections; it represents the extent to which dense cohesive nodes in a network are interconnected (Pryke, 2004).

As this is a weighted network, the weighted degree must be represented. The evaluated types of centrality are ‘betweenness centrality’ and ‘closeness centrality’ (Freeman, 1979). Betweenness centrality signifies the extent to which a node lies between other node pairs. Closeness centrality indicates the ability of a node to access information through other nodes.

The results of the first two steps of this review are presented in Table 2, which shows the incidence of the keyword ‘principal–agent’ in the seven leading construction management journals. Overall, 148 papers containing the main keyword were found; the main keyword appeared in four paper titles, 22 abstracts and in no more than 14 keyword lists, which were essential in the literature search. The main keyword appeared in IJPM the most (70 papers), followed by JCEM (25 papers), CME (23 papers), ECAM (17 papers), CI (7 papers), BRI (5 papers) and AIC (1 paper).

The 148 papers cited in the top seven construction management journals are presented in Table 3. Most of the papers were published in the 2000s, mostly between 2005 and 2020. The peak was in 2019; 14 papers with the keyword ‘principal–agent’ were published. In recent years, the numbers have been similar: 13 papers in 2014, 11 in 2015, 13 in 2016 and 12 in 2017 and 2018. Most of the research has been recent. However, the first papers appeared in 1991 (Ward et al., 1991; Ward and Chapman, 1991).

The associated keywords from 148 selected papers were classified and analysed to determine their connection to the main keyword ‘principal–agent’; they reflect the themes of the papers.

The analysis identified 717 keywords, which amounts to less than five keywords per paper. Several papers shared a large number of identical keywords. Owing to space limitations, only the most important are presented here. Keywords such as ‘construction’ and ‘construction industry’, ‘construction management’ and ‘project management’ were excluded from further analysis. General keywords such as ‘environment’ and ‘sustainability’, and indefinite keywords such as ‘control’, ‘conflict’, ‘flexibility’ and ‘performance’ were also excluded. Names of countries and organisations were excluded, as were technical terms such as ‘Monte Carlo method’ and ‘regression analysis’. The main keyword ‘principal–agent’ was also excluded because the 14 papers in which it explicitly occurred were already considered. The incidence of the 36 remaining associated keywords in the seven selected journals is shown in Table 4. They appear 209 times – 70 times in IJPM, 54 times in JCEM, 43 times in CME, 24 times in ECAM, nine times in CI, seven times in BRI and twice in AIC.

Visual representations of keywords from Table 4 and their interconnections are shown in Figure 1. Four nodes have been omitted from the visual representation because they do not have connections with other keywords: ‘change management’, ‘corruption’, ‘social theories’ and ‘temporary organisations’. They have a connection with P–A theory somewhere in the article text, but not in the keywords section. Detailed visualisation data are presented in the Appendix. The visualisation of the network was created using the Force Atlas layout. Nodes are modelled with a larger diameter to indicate a larger weight, and a darker colour if they are more central (by betweenness centrality). The specific node characteristics are listed in Table 5. Edges are modelled to be wider and darker according to their weights.

Fig. 1

The graph density was calculated to evaluate the overall keyword connectivity. A density of 0.188 indicates weak interconnectedness; a small portion of potential connections are connected. A total of 33 nodes were analysed; the average degree was 6 and the average weighted degree was 10.24.

The importance of each node was evaluated by considering the degree and weighted degree. The ten nodes with the largest degree were ‘contracts’ (18), ‘P–A theory’ (16), ‘partnership’ (14), ‘incentives’ (13), ‘governance’ (12), ‘risk management’ (11), ‘trust’ (11), ‘transaction costs’ (10), ‘information systems’ (9) and ‘procurement’ (9). Considering the weighted degree, ‘contracts’ have the most information transactions in the analysed network (46 connections), followed by ‘P–A theory’ (28), ‘partnership’ (28), ‘incentives’ (27), ‘governance’ (24), ‘risk management’ (24), ‘transaction costs’ (23), ‘trust’ (14), ‘agency theory’ (12) and ‘procurement’ (11).

The centrality of the node was evaluated through closeness and betweenness centrality. Centrality measures the social power and influence of a node. Much information flows between most keywords, but their ability to access information through other nodes is not strong, indicated by a closeness centrality of approximately 0.5 for 29 analysed keywords.

According to betweenness centrality, ‘P–A theory’ (68.26) and ‘contracts’ (61.63) are by far the most important elements in the network, followed by ‘governance’ (40.0), ‘partnership’ (39.53), ‘transaction costs’ (39.07), ‘information systems’ (37.42), ‘incentives’ (30.26), ‘risk management’ (15.92) and ‘trust’ (15.62). The other nodes have little or no centrality. Thus, the majority of information flowing in the network is controlled through the keywords ‘P–A theory’ and ‘contracts’. The removal of these nodes from the network may fragment the network (Home Office, 2016). The centrality of ‘P–A theory’ was expected because it was the central keyword in the literature search. The centrality of ‘contracts’ can be explained by its importance in the construction industry. Since the keyword ‘contracts’ has a much greater weighted degree, it can be concluded that this is the most important keyword in the network.

Sometimes the analysed keywords were mentioned by themselves in the keywords of a paper. This is especially true for ‘governance’, which was researched ‘on its own’ in nine papers (see Appendix). The isolation and a low graph density clearly indicate a low connectivity with the theoretical concepts of P–A theory in construction management research.

The literature indicates that the main research interests associated with P–A theory in construction management are contracts, partnership, incentives, governance, risk management, transaction costs, trust and procurement. A main finding of this study is that the most important (central) elements associated with P–A theory in construction management are contracts, governance, partnership, transaction costs, information systems, incentives, risk management and trust.

These associated keywords are the focus of recent research. Nevertheless, some of the keywords are often researched without considering related theoretical concepts. For example, ‘governance’ is only loosely connected with related concepts in a significant number of the retrieved papers.

Construction procurement is defined by contracts. Many participants may be involved in the delivery of a project. There may be several contracts, sometimes in different forms. The main aim of contracting is to properly allocate risk between the project owner and the contractor. This is applicable to any P–A relationship in the project. For proper risk allocation, the sharing of project outcomes is facilitated by incentives, which are an important vehicle to align contractor interests with owner interests. Optimal incentive contracts have been a research focus in recent years (Hosseinian and Carmichael, 2013). Choosing the best incentive strategy is critical in construction procurement (Chang, 2014).

Nevertheless, contracts are not sufficient to prevent negative opportunistic behaviour by participants. The main problem is that in any project, a large number of participants suggest a large number of non-contractual relationships between them (Cerić, 2016). Thus, in addition to formal methods of control (written contracts), researchers are now exploring how relational contracting can minimise project risks and transaction costs. Partnership is common in construction projects because it enhances trust between partnering organisations, which reduces risk. Trust is an important part of relational contracting; an individual or organisation trusts another individual or organisation to do what they have promised (Benítez-Ávila et al., 2018). Trust is also known as a governance mechanism or an informal control mechanism. Thus, project governance is associated with the project structure and formal contracts, and also with relational contracting, focusing on P–A theory and communication risks (Ahola et al., 2014). Information systems such as BIM and blockchain can influence the development of trust and cooperation among project participants. Thus, they are another focus of recent research on minimising communication risks (Cerić, 2019).

Transaction cost theory is a part of New Institutional Economics, a theoretical framework that supports P–A theory. It emphasises the economic importance of creating or selecting governance structures for individual transactions to reduce contractual hazards (De Schepper et al., 2015). Transaction cost theory is based on the idea that major transaction costs can arise from asymmetric information and opportunistic behaviour. Bromiley and Cummings (1995) claim that trust reduces transaction cost and suggest that the inclusion of trust would expand and extend the research framework of transaction cost economics.

Other interesting keywords were also found; however, the discussion has focused on the most frequent and central keywords because they represent the bulk of current research.

Construction projects are delivered through a multitude of participants or stakeholders; between them, there are different contractual and non-contractual relationships. The complex relationships between project participants along with individual self-interest provide a fertile ground for opportunistic behaviour and common communication risks. P–A theory and related concepts can complement our understanding of construction management.

The purpose of this paper was to assess new developments in construction management by analysing the incidence of P–A theory in construction management literature. A literature analysis was conducted using keywords listed by the authors. The analysis identified eight research concepts currently of the greatest interest (contracts, partnership, incentives, governance, risk management, transaction costs, trust and procurement) and centrality (contracts, governance, partnership, transaction costs, information systems, incentives, risk management and trust) in construction management papers.

Considering the framework of New Institutional Economics, this review does not cover all related theoretical concepts. The overlap of theories within the framework of New Institutional Economics, P–A theory, and transaction cost theory is worthy of future study.

Although this review was intended to provide a comprehensive image of P–A theory trends in the construction management literature, it has limitations in terms of journal selection. The archives of the selected journals span different years. The BRI archive goes back to 1973; CME, JCEM, and IJPM go back to 1983; AIC goes back to 1992, ECAM goes back to 1994 and CI goes back to 2001. Thus, journals with older archives are represented to a greater extent in this review. The bulk of the cited literature was published within the past two decades. Thus, the historical reach of the archives does not appear to be a factor.

The selected journals were identified by Bröchner and Björk (2008). Since 2008, many new journals have appeared, and are gradually becoming more important. As a result of the sampling decision, certain relevant publications may not have been included in this review, although the seven journals selected for investigation are the best. The literature review should be extended to a wider selection of journals to provide further understanding of the field.

Opportunistic behaviour, self-interest, corruption, communication risks, trust and other issues within the focus of P–A theory are widespread throughout the construction industry. It is hoped that this review offers a new perspective to construction management researchers, and that they can find new areas of interest within the list of associated keywords and interconnections identified by NA.