The Relationship between Building Agents in the Context of Integrated Project Management: A Prospective Analysis

: The increasing complexity of building projects, with high quality standards, integrated technologies and strong management restrictions, demands the intervention of numerous and diverse specialists. This requires an intense leadership, organization and coordination effort. However, building regulations, such as the Spanish Law on Building Management (LOE) 38/1999, only formally consider developers, project designers, project and work supervisors, quality control entities, construction companies, owners and ﬁnal users as building agents. However, these categories are insufﬁcient to represent the interests of all the stakeholders. The purpose of this paper is to carry out an analysis of the agents that are currently part of the entire building process. If their relationship of inﬂuence and dependence, as well as their alignment with the overall objectives of the project are studied, potential convergences, divergences, agreements and disagreements can be established. To do this, the authors conducted a prospective analysis through the MACTOR (Matrix of Alliances and Conﬂicts: Tactics, Objectives and Recommendations) strategic planning simulation tool, for which the rules of the Delphi technique were applied and a consultation with technical experts, both professionals and academics, was held. The research provides insight to assess the power relationships between the building agents, as well as to measure the alignment of objectives with their interests. Results show that, in the context of integrated project management (IPM), the inﬂuence of technical agents is reduced by limiting their functions to those marked by their regulatory framework, allowing them to focus on their legal powers, and the room for manoeuvre of the professional agents, who are subject to systematized monitoring and control, is also reduced. The prospective analysis also highlights the importance of deﬁning the scope from its early stages, as well as the need to reach multilateral agreements based on the other two main constraints: time and cost.


Introduction
Despite the poor performance of construction industry projects [1], their contribution to the development of the international economy over the past 40 years has been crucial, in Africa [2], America [3], Asia [4], Europe [5] and Oceania [6]. In Spain, this industry has exemplified the intense and complex crisis the country has suffered over the period from 2008 to 2014 [7], commonly referred to as the Spanish real estate bubble [8].
The performance of construction projects is influenced by many aspects [9], both technical (limitations, constraints and assumptions, risks and opportunities, etc.) and relational (changes, stakeholders' needs and expectations, etc.). If those involved in the project state their demands at a very early stage [10], then the requirements to be considered in the project definition can be identified early on, which greatly increases the chance of the project completing successfully [11]. Acceptance or rejection of these requests may be

•
Control entities and sponsors, such as financial institutions and insurance companies [43], as well as self-regulatory professional bodies, which in Spain are professional associations that check building requirements [44]. • Technicians, such as consultants of topographical and geotechnical studies, partial designers of technical projects (or subcontractors of project designers), interior designers, landscapers and decorators [45], as well as partial supervisors of work construction (or subcontractors of work directors).

•
Managerial agents, such as site foremen [46] (who traditionally represented the contractor but are being independently incorporated as construction managers [47]) and waste managers [48] (dealing with recycling, valorisation and disposal). • Specialised professionals [49], such as construction subcontractors.
After the publication of the LOE in 1999, the construction sector became one of the drivers of the Spanish economy [50], due to its contribution to gross domestic product (GDP), its capacity to generate jobs and the multiplying potential of other auxiliary industries, until the bursting of the real estate bubble [51], when it halved its share of national GDP from 10.41% in 2008 to 5.22% in 2014. In addition, according to data from the National Statistics Institute (INE), in these last 20 years, although productivity in civil engineering works has almost doubled (up 88.27%), in building it has barely increased by 1.74% [52].
The reality of the building sector in Spain led to the incorporation of new contracting formulas beyond the traditional contract, with the aim of optimising investment, applying technological advances and ensuring the construction and delivery. Among these organisational models, those which stand out are the project and construction (turnkey) model [53] and the IPM [41]. Other contract types less used in Spain are the accelerated construction process, the build-own-operate-transfer (BOOT) model, the engineering model and open book pricing, among others.
The traditional model is one in which, on the one hand, the developer contracts the architectural and/or engineering services to design the technical project and assume the functions and responsibilities of the supervision of the work, and, on the other hand, formalises a construction contract with the construction company for the material execution of the work and coordination of subcontractors. In this way, the whole process is divided into the celebration of two non-concurrent main contracts:

1.
Between the developer and the technicians, by which they undertake to define the technical specifications of the building work and even prepare the tender, advising on the choice of the construction company, as well as supervising the execution of the building work.

2.
Between the developer and the construction company, by means of which, the latter is obliged to the developer to carry out the work in accordance with the technical project. As the technical project was designed prior to the tender, this promotes competitiveness among bidders, reducing the problems of a lack of definition or

Prospective Analysis
How can the power relationships of each building agent be determined? What is this relationship between interests? Is it possible to measure these interactions? How can they be measured? How do the positions of the stakeholders within the project impact its outcome? To answer these research questions, a prospective analysis has been done. Prospective analysis is a qualitative method that is based on action and non-predetermination using specific methods for setting out a range of future scenarios [58]. This methodology includes a stakeholder analysis of the parties involved in the system to be studied. To undertake this stakeholder analysis, the MACTOR strategic planning simulation tool is used [59] because of its facilitation and systematisation capacity. This tool is designed to assess the power relationships between the project stakeholders, as well as to measure the alignment of objectives from the associated particular interests [60].
The MACTOR method seeks to gauge the balance of power between stakeholders and study their convergences and divergences when faced with a certain number of associated objectives, allowing them to determine alliances and coalitions and avoid conflicts [61]. This tool provides an analysis to predict the future behaviour of stakeholders that operate in a project: What is their position in relation to this or that problem? With whom could alliances be formed? How strong are they? How reliable are they? With whom will it be more difficult to negotiate? These are some of the major uncertainties addressed by the MACTOR tool.
The MACTOR analysis requires a sequence of steps [71] that can be described in these three stages: 1.
The building agents (parties actively involved) as well as the strategic objectives of the entire building process are identified.

2.
The data are entered into the software, converting qualitative data in accordance with prescribed procedures, with which the results are obtained: • Identification of the strategic stakes and associated goals and of the position of each actor according to each objective.

3.
The data are analysed to make recommendations or proposals to the building agents, based on their relationships and positions regarding the project objectives (for example, formulating strategic recommendations and asking key questions for the future).
As the sequence of steps progresses, MACTOR produces tables and graphs. These results obtained will provide a clearer picture of how the system actually works. However, it is necessary to define one step prior to carrying out this method: to establish the data sources which will feed the system. These data may come from content analysis, literature review, seminars, workshops, questionnaires and interviews with stakeholders or experts [72]. In this research, results will come from a panel of experts, using the Delphi technique. Identifying and selecting suitable opinion experts is crucial for the success of the research [73], becoming one of the most critical actions in the whole process [46]. Although these experts can have differentiated decision-making habits, risk preferences or knowledge backgrounds [74], the useful guidance provided by them indicates that they are confident in the field [75]. Once the process starts, the experts arrangement through the Delphi Technique allows the level of consensus and the hierarchy of its importance [76] to be determined.

Selection of Experts
The Delphi technique has to be executed with the participation of individuals who have knowledge and competence in the study subject, as well as a deep understanding of the problem [77]. Accordingly, to be part of the initial sample, it is necessary for the panel members to be building experts. To do this, initial requirements included having both relevant experience and deep knowledge on the subject (building life cycle, building project management, construction management, construction engineering, legal regulatory framework, real estate promotion, etc.), so academics who are or have been professionals, and professionals who are or have been academics were sought. In addition, in order to avoid bias, candidates had to meet four additional requirements:

1.
Expertise, either in building project management or in construction management, or in designing technical projects or directing works. Ten years of experience are required, having participated in at least five projects of more than one million euros each. The possession of professional project management certifications is positively valued.

2.
Relevant teaching experience, with at least another ten years of acknowledgement (two recognized five-year periods) in this field. 3.
Pioneering research, with at least twelve years of impactful research (two recognized six-year periods) in this field.

4.
Transfer experience, with at least five collaboration contracts in this field with companies for applying research findings in a specific context.

Identification of Agents and Strategic Objectives
Once the panel of experts is formalized, a preliminary list is extracted from the current legal framework for the identification of agents. This is distributed to the panel of experts to suggest changes and comments, indicating that the research is conducted in the context of a IPM contract comprised of the technical project and the building work. This took two rounds (after a first round in which some actors not explicitly included in the regulation were proposed and some clarifications were made) to reach an agreement. Table 1 shows this final list, including the main phases of the process in which the building agents participate. On the contrary, a preliminary list of strategic objectives is extracted from the literature review of the last two decades. This is distributed to the panel of experts to suggest new references. Although no objectives were added or removed, some references were included, as shown in Table 2. Classic managerial objectives such as cost, time, quality and scope (of work) [78][79][80][81] are joined by instrumental objectives such as health and safety and environment [82][83][84][85][86] and communication [87,88], technical and professional objectives such as functionality, productivity and technical performance [89,90], and client objectives such as profitability [91,92] and customer satisfaction [93,94]. From Tables 1 and 2, both questionnaires are sent to the expert panel.   Decade  Source  O01  O02  O03  O04  O05  O06  O07  O08  O09  O10  O11  O12   C  S  Q  SW  HS  E  F  P  TP  PF  CM  CS

Data Collection
In relation to the agents, following the instructions of the MACTOR tool, a hierarchical classification is carried out using the matrix of direct influences (MDI), quantifying the direct influences (DI) and dependences (DD) between agents, in accordance with a scale of 0-4: (0) if agent Ai cannot question agent Aj, (1) if agent Ai can question agent Aj's operations (limited in scope and time), (2) if agent Ai can question the choice of agent Aj, (3) if agent Ai can question agent Aj's roles, and (4) if agent Ai can question the existence of agent Aj.
Regarding the objectives, there is to determine the position of agents in relation to each objective, which is represented in the valued positions matrix (2MAO). To do this, first it must be determined whether the agent is favourable or unfavourable towards the objective: (+) if agent Aj is in favour of objective Oi, (−) if agent Aj is against objective Oi and (0) if agent Aj is neutral towards objective Oi (this represents the simple positions matrix (1MAO)). In second place, the intensity of the agent's position on that objective is measured: (0) if objective Oi has little or no impact, (1) if objective Oi can question (limited in scope and time) the operations of agent Aj or is essential to them, (2) if objective Oi can question the choice of agent Aj or is essential for it, (3) if objective Oi can question the performance of agent Aj's duties or is essential for them, and (4) if objective Oi can question the existence of agent Aj or is essential for it.
Experts must consider each agent's plans, motivations, limitations and means of action. The estimations of the experts are made in successive anonymous rounds, with maximum autonomy on the part of the participants. The experts are asked to include appropriate comments to justify their answers, so that the other experts can reason this out in subsequent rounds. The consultation process ends once an agreement is reached. In order to obtain the MDI and 2MAO, this agreement is achieved through the consensus among panel members (convergence of opinions expressed) and the stability of their responses (limited variation in the responses given by the group). On the one hand, consensus is obtained across the interquartile range (IQR), by calculating the difference between the third quartile (Q3) and the first (Q1), with a variation equal to or less than 0.5 being accepted [97]. On the other hand, stability is calculated using the relative interquartile range (RIR), which results from dividing the IQR by the second quartile (Q2), and is accepted if the variation is less than 0.3 [98]. It can be noted that the values entered in MACTOR represent the median (Q2) for each cell value, once it is verified that both the IQR and RIR do not exceed the maximum permitted values for each row i and column j. Then, the software subsequently evaluates the power relationships between them.

Relationships between Agents
If the influences that each agent has on the others are added together, without considering the indirect self-influences, the matrix of direct and indirect influences (MDII) is obtained, which shows the direct and indirect influences and dependences (II, ID) of second order between agents [60]. Next, using the MDVI, the competitiveness relationship (Ri) between agents is also determined, considering both influences and dependences and their retroaction. These relationships ascend as their influence increases (and/or their dependence decreases) and vice versa, anticipating the pre-eminence of one agent over another. Then, the net scale (NS) of influences is determined, measuring for each pair of agents the distance between DI and II, indicating for each pair of agents the surplus influence exerted or received (which translates into the most and least influential agents). After that, the influence-dependence graph between agents is obtained. This graph shows the degree of dependence on the X-axis and the impact of influence on the Y-axis. Thus, agents are classified into four types: dominant (very influential, not very dependent), intermediary (influential, but also dependent), autonomous (not very influential and independent) and dominated (not very influential, very dependent). Finally, the maximum level of influence that an agent can exert on another agent, directly or through a intermediary agent, is identified, and the matrix of maximum direct and indirect influences (MMDII) is obtained. From the MMDII, the maximum competitiveness ratios (Qi) are determined, summarizing in a single value the degrees of maximum influence and dependence of each agent, giving a measure of their real relationships [99]. This process is summarized in Figure 1: Buildings 2021, 11, x FOR PEER REVIEW 9 of 39 panel members (convergence of opinions expressed) and the stability of their responses (limited variation in the responses given by the group). On the one hand, consensus is obtained across the interquartile range (IQR), by calculating the difference between the third quartile (Q3) and the first (Q1), with a variation equal to or less than 0.5 being accepted [97]. On the other hand, stability is calculated using the relative interquartile range (RIR), which results from dividing the IQR by the second quartile (Q2), and is accepted if the variation is less than 0.3 [98]. It can be noted that the values entered in MACTOR represent the median (Q2) for each cell value, once it is verified that both the IQR and RIR do not exceed the maximum permitted values for each row i and column j. Then, the software subsequently evaluates the power relationships between them.

Relationships between Agents
If the influences that each agent has on the others are added together, without considering the indirect self-influences, the matrix of direct and indirect influences (MDII) is obtained, which shows the direct and indirect influences and dependences (II, ID) of second order between agents [60]. Next, using the MDVI, the competitiveness relationship (Ri) between agents is also determined, considering both influences and dependences and their retroaction. These relationships ascend as their influence increases (and/or their dependence decreases) and vice versa, anticipating the pre-eminence of one agent over another. Then, the net scale (NS) of influences is determined, measuring for each pair of agents the distance between DI and II, indicating for each pair of agents the surplus influence exerted or received (which translates into the most and least influential agents). After that, the influence-dependence graph between agents is obtained. This graph shows the degree of dependence on the X-axis and the impact of influence on the Y-axis. Thus, agents are classified into four types: dominant (very influential, not very dependent), intermediary (influential, but also dependent), autonomous (not very influential and independent) and dominated (not very influential, very dependent). Finally, the maximum level of influence that an agent can exert on another agent, directly or through a intermediary agent, is identified, and the matrix of maximum direct and indirect influences (MMDII) is obtained. From the MMDII, the maximum competitiveness ratios (Qi) are determined, summarizing in a single value the degrees of maximum influence and dependence of each agent, giving a measure of their real relationships [99]. This process is summarized in Figure 1:

Alignment of Agents and Objectives
The 2MAO matrix quantifies the agents' positions on the objectives.

Alignment of Agents and Objectives
The 2MAO matrix quantifies the agents' positions on the objectives. From this, the simple position matrix (1MAO) is obtained, focused on the agents' assessments, indicating each agent's position on each objective (for, against or neutral). If the values of the MIDI and the 2MAO are combined, the valued weighted positions matrix is obtained (3MAO), which considers the degree of opinion of each agent on each objective, hierarchy and competitiveness. Positive values represent the acquiescence of agents in achieving their objectives and negative ones represent their opposition to achieving them. Each of the three matrices provides the number of agreements (OK) and disagreements (KO) of the agents with the objectives and the convergences (CV) and divergences (DV) between agents, as well as their mobilisation (MB). Once these matrices are calculated, the net distances between agents can be obtained, which allows potential alliances to be recognized, as well as the net distances between objectives. The conjunction of these two allows shared and contrary positions to be determined according to agent and objective, enabling these common agreements to be worked on [99]. This process is summarized in Figure 2: which considers the degree of opinion of each agent on each objective, hierarchy and competitiveness. Positive values represent the acquiescence of agents in achieving their objectives and negative ones represent their opposition to achieving them. Each of the three matrices provides the number of agreements (OK) and disagreements (KO) of the agents with the objectives and the convergences (CV) and divergences (DV) between agents, as well as their mobilisation (MB). Once these matrices are calculated, the net distances between agents can be obtained, which allows potential alliances to be recognized, as well as the net distances between objectives. The conjunction of these two allows shared and contrary positions to be determined according to agent and objective, enabling these common agreements to be worked on [99]. This process is summarized in Figure 2:

Results
A total of eighteen respondents finally participated from 15/04/2020 to 14/10/2020, after accepting the proposal for participation and meeting the requirements established to be considered experts. The invited group of experts came from six areas of knowledge (Architectural Buildings, Administrative Law, Civil Law, Construction Engineering, Architectural Projects and Engineering Projects) from eight Spanish universities, accumulating an average of forty-one building projects (from six to one hundred and thirty-nine projects), five five-year periods of teaching experience and four six-year periods of research (from thirteen to forty-seven years), and sixteen transfer contracts (from five to thirty-six), as is shown in Table 3. In summary, the experts have extensive experience in building-related research and teaching, as well as active participation in numerous construction projects, demonstrating their expertise in the generation, dissemination and transfer of knowledge in the field.
Once the group of experts was obtained, they proceeded to weigh the direct relationships between agents (getting the MDI) and of these agents with the general objectives of the project (getting the 2MAO). To this end, they are provided with the final list of the building agents and the overall objectives, so that these can be assessed, according to the scales defined. In the case of stakeholders' relationships of power, the experts reached an agreement after the third round in the MDI (with an IQR of 0.42 and a RIR of 0.24). In the case of the position of stakeholders with regard to the project objectives, the experts reached an agreement in the fourth round in the 2MAO (with an IQR of 0.48 and a RIR of 0.28). The resulting final matrices are shown in Tables 4 and 5.
After consensus and stability were achieved in the expert panel consultation, the matrices' data (MDI, 2MAO) were loaded into the MACTOR software (version 5.1.2). Through this tool, the results processed were subsequently analysed, establishing hierarchies of the relationships between the agents, as well as the relationships of the agents

Results
A total of eighteen respondents finally participated from 15/04/2020 to 14/10/2020, after accepting the proposal for participation and meeting the requirements established to be considered experts. The invited group of experts came from six areas of knowledge (Architectural Buildings, Administrative Law, Civil Law, Construction Engineering, Architectural Projects and Engineering Projects) from eight Spanish universities, accumulating an average of forty-one building projects (from six to one hundred and thirty-nine projects), five five-year periods of teaching experience and four six-year periods of research (from thirteen to forty-seven years), and sixteen transfer contracts (from five to thirty-six), as is shown in Table 3. In summary, the experts have extensive experience in building-related research and teaching, as well as active participation in numerous construction projects, demonstrating their expertise in the generation, dissemination and transfer of knowledge in the field. Once the group of experts was obtained, they proceeded to weigh the direct relationships between agents (getting the MDI) and of these agents with the general objectives of the project (getting the 2MAO). To this end, they are provided with the final list of the building agents and the overall objectives, so that these can be assessed, according to the scales defined. In the case of stakeholders' relationships of power, the experts reached an agreement after the third round in the MDI (with an IQR of 0.42 and a RIR of 0.24). In the case of the position of stakeholders with regard to the project objectives, the experts reached an agreement in the fourth round in the 2MAO (with an IQR of 0.48 and a RIR of 0.28). The resulting final matrices are shown in Tables 4 and 5. After consensus and stability were achieved in the expert panel consultation, the matrices' data (MDI, 2MAO) were loaded into the MACTOR software (version 5.1.2). Through this tool, the results processed were subsequently analysed, establishing hierarchies of the relationships between the agents, as well as the relationships of the agents with the project objectives. This information is of great relevance, since from it the prospective analysis of building agents is generated. Table 6 summarizes both direct, indirect and maximum influences and dependences between agents: DI, DD, II, ID, MI, MD. Directly, the great influence wielded by project managers with 53 points (as managerial leaders during the whole process), the public administrations and the professional associations with 17 points (as controllers of legal and technical requirements), project designers with 26 points (as technical leaders during the technical project) and project and work supervisors with 16 points (as technical leaders during the building work) can be highlighted. However, if indirect influences are also considered, although project managers remain the most influential agents (with 226), project designers (with 102) and project and work supervisors (with 125) lose their position to public administrations (with 157) and professional boards (with 128). If maximum influences are considered, project managers (with 75), project designers (with 36), public administrations (with 32) and project and work supervisors (with 30) must be stressed. In terms of dependence, the position of professional associations, project designers and construction companies stand out, both in direct and indirect terms. The entire MDII and MMDII matrices are shown in Tables A1 and A3 (see Appendix A).
The sign indicates whether the actor is likely to reach objective or not. 0: Objective has a bleak outcome.   Table A2) collects the net scale: NS of influences for each agent. In this case, the net influence of project managers (with 209 points), public administrations (with 157), construction managers (with 84) and project and work supervisors (with 81) stands out, as well as the net dependence of product suppliers (with 82 negative points), self-employed workers (with −73) and even construction companies (with −69). Additionally, Table 8 draws the competitiveness (and maximum competitiveness): Ri, Qi in each agent's case, in which the positions of project managers (by delegation of the developers), real estate developers (as direct clients) and public administrations (for their power to demand legally compliant solutions), as well as construction managers (as assistant project managers during the construction phase) and owners (due to exploitation of the promotion) highlight above the rest. The project designers are in an intermediate position, even though they have the intellectual right to the object of the construction. Likewise, those responsible for ensuring that the work is carried out in accordance with the contracts and the rules for smooth construction (project and work supervisors and control entities) are also at an intermediate level of the hierarchy. On the other hand, the actors in charge of materialisation (main and specific contractors) are at the bottom of the scale and are highly dependent on the other agents involved. Table 9 summarizes simple, valued and weighted valued of convergences and divergences: CV, DV between agents (see the entire matrices in Tables A4-A11). Both in terms of alignment and mobilisation, the position of project managers and construction managers stands out for their contribution to the leadership, organization and coordination of the process. In addition, project managers are also highlighted in terms of divergence, along with the real estate developers, who are the initiators of the entire building process and recruiters directly or indirectly of most of the participants. Additionally, it is worth noting the neutral position of public administrations and professional associations. Finally, Table 10 summarises the simple, valued and weighted valued of agreements and disagreements: OK, KO, as well as the mobilisation: MB that the objectives cause between the agents. The objective that achieves the most agreements is the project scope (with 31, 104 and 90.9 points, respectively), followed by the project quality (with 31, 88 and 71.4) and the technical performance (with 28.77 and 74.3), all of them related to compliance with requirements. On the contrary, the objectives that cause the most disagreements among agents are the project cost and time (with −27, −73 and −67 points), the other two elements of the "triple constraint", followed by its profitability (with −25, −47 and −30). Figure 3 shows the map of relationships between agents, differentiating between the technical project and the building work. In both stages, the dominant position of project managers and public administrations stands out, whereas the construction managers do so during the building work. In addition, real estate developers, owners, end-users, financial institutions, insurance companies, control entities and waste managers also act as autonomous agents throughout the whole process. During the technical project, the main figures recognised as building agents in the LOE (project designers and telecommunications designers) act as intermediary agents (together with health and safety coordinators and professional associations), and those outsourced by them (structural designers, facility designers, energy efficiency certifiers and project health and safety coordinators), as dominated agents (together with environmental consultants, health and safety planners and use and maintenance planners). In contrast, owners, end-users, financial institutions, insurance companies, and control entities do so as autonomous agents, as well as topographical and geotechnical consultants and interior designers, landscapers and decorators. During the building work, those figures recognised as building agents in the LOE (work directors, executive work directors, telecommunications supervisors and work health and safety coordinators) also act as intermediary agents, and those outsourced by them (structural and facility supervisors), construction companies and their entire supply chain become dominated agents. ID 41 55 119 92 103 72 85 26 99 92 89 90 62 19 18 17  0  181 19 20 20 81 44 64 74 75 81 78 105 9  19 44 99 108 MD 29 31 32 31 31 30 30 25 30 27 30 30 20 11 14 13  0  39 12 12 13 29 20 29 29 29 29 27 22  9 19 19 20 26 Table 7. Summary of Net Scale of Influences and Dependencies between Actors (NS). Extracted from Table A2.       Figure 4 shows the net distances between agents in relation to their positions on the project objectives. During the technical project, it is worth noting the distance between the technical agents explicitly listed in the LOE and their subcontractors (who may also act as partial technical agents, hence their conflicting positions). On the other hand, during the building work, it is worth noting the distance between the construction managers and the construction companies and their entire supply chain (thanks to this position, directors and supervisors can focus on the fulfilment of requirements). However, both in the technical project and in the building work, the aligned position of real estate developers, owners, end-users (as direct and indirect clients) and project managers stands out. Finally, Figure 5 shows the net distances between objectives. During the design of the technical project, it is worth noting the distance between the scope and quality (stressing the difference between doing the right things and doing the things right) and between cost and time (as the reverse side of the same purpose), as well as the alignment between health and safety and environment. On the other hand, during the building work, it is worth noting both the distance between scope and quality, and cost and time (as in the design of the technical project). On the contrary, throughout the whole process, the alignment of profitability, productivity and technical performance stands out.

Discussion
Scenario analysis based on prospective tools [99] has proven useful in preparing possible potential scenarios. This is one of the crucial stages of planning [100], as the resolution of conflicts stemming from having differing individual goals conditions the evolution of the project. In this setting, once the relevant agents in a IPM context are identified, a stakeholder analysis has made it possible to understand their power relationships and alignments. This improves stakeholder management, which is a key factor in achieving optimal outcomes [101]. Nevertheless, this would require consistent behaviour on the part of agents in achieving their goals, which may not happen, as they are reluctant to reveal their strategic approaches and their means of action. Therefore, instead of using the stakeholders themselves as a method for gathering the data [102], a panel of experts using the Delphi Technique [103] is established, with a notably neutral nature. If reliable information is available, the needs of stakeholders relating to information [104] can be grouped by their alignments, agreements and convergences.
From a theoretical point of view, the results of the MACTOR analysis explain that the influence and interdependence between stakeholders show the strength of the competitiveness of stakeholders in determining the objectives to be formulated. The higher the competitiveness, the stronger the influence in accommodating each other's interests. The MACTOR score also indicates the level of stakeholder approval. In addition, this score shows the convergence and divergence between stakeholders and compiles their agreements and disagreements. Thanks to the measurement of the relationships of power of each pair of stakeholders and the position of each stakeholder with respect to each objective provided by the panel of experts, the MACTOR tool allows the:

•
Creation of agents' strategy, once their goals, objectives, preferences, motivations, constraints, coherence and attitude are identified. • Identification of strategic stakes and their corresponding objectives, once converging and diverging objectives between agents are determined.

•
Position of each agent with respect to the strategic objectives.

•
Gathering of all possible tactics (alliance and conflict games), once the stakeholders' position with respect to the different objectives is evaluated.

•
Calculation of the competitiveness between stakeholders, once the structure of direct and indirect influences is analysed.

•
Integration of the competitiveness of each stakeholder to the degree of his/her/its positioning with respect to objectives. • Formulation of strategic recommendations and key questions for the future, which are beyond the scope of this research.
From a practical point of view, the network of influences and dependences between the agents and their position regarding the overall objectives allows potential conflicts and alliances to be identified, thanks to their agreements, disagreements, convergences and divergences. Thanks to prospective analysis, the direct and indirect influences between each pair of agents, as well as their dependences, are known. Once the agreements, disagreements, convergences and divergences of these agents with respect to each project objective are known, alliances can be established and conflicts avoided. This can apply to different regulatory frameworks and different contract modalities between the agents involved in the building process, thanks to the MDI (which measures the jeopardy between each pair of agents) and 2MAO (which measures the intensity of the priority of each objective for each agent) matrices. Different contexts will lead to different relationships of power between agents and different positions of the agents with respect to the project objectives, but thanks to MACTOR it will be possible to gather all possible tactics (alliance and conflict games). In summary, this tool enables the visualisation of agent groups (with converging interests), their apparent degree of freedom, also to identify those agents potentially more at risk and to analyse the stability of the system. IPM contracts generate multiple benefits in organisations, reducing the failure rate of projects undertaken [105], so that traditional contracts included in the Spanish LOE are being superseded. Despite not being explicitly included in the Spanish legal system, the figure of the project manager (for the entire building process) and of the construction manager (for the contracting and building work), must be highlighted for their central roles. In this study, both project and construction managers achieve the highest number of convergences, establishing the greatest number of relationships with the other agents. In this way, this study can be used as a precedent in Spain to address the responsibilities of these agents, placing them fully in the building process, despite having no specific professional powers reserved [41]. However, all the agents must reflect their respective interests in defining the project from the beginning [10], for which they must know the hierarchical structure of their relationships, specially project designers and work supervisors, which are the following agents in number of convergences and play the managerial roles in traditional contracts [106]. Accordingly, the IPM places the Spanish system in an international context, mitigating the effects of the singularity of the Spanish regulatory framework [44] related to the roles, functions and responsibilities of the active stakeholders (buildings agents) that are present during the process. On the contrary, the neutral position of final users (because they are not present during the building process), public administrations and self-regulatory professional bodies can be stressed.
With regard to the position of each agent with respect to the project objectives, in this study the scope objective has achieved the most agreements among the building agents identified (90.9%). This result is aligned with the literature that holds that most of the conflicts and crises between building agents that occur in a building project stem from a lack of (or mis-) definition [107]. This lack of definition can lead to moments of tension when responsibility is sought from the rest of the agents. In fact, the lack of cooperation between agents has been identified for almost 20 years as one of the biggest causes of deficiency in the construction industry [108]. In this scenario, knowing about their relationships of power is critical to finding a solution to these problems. In addition, the quality and the technical performance objectives are the second (76.4%) and third (74.3%) most agreed objectives, respectively, confirming that legal and stakeholders' requirements must be considered, but doing it in an effective and efficient manner. This contributes to client satisfaction [109], the fourth objective with the most agreement among the agents in this study (61.5%). The following objective to be taken into account was productivity (57.8%). This contrasts with the lack of productivity in the Spanish building sector, which has barely increased by 1% in the last 20 years, despite the increase in the turnover of Spanish construction companies in recent years. The rest of the objectives do not reach 50% agreement, although health and safety and environment are close (44.7% and 46.4%, respectively). Functionality (28.9%), communication (28.3%) and profitability (23.0%) are only around 25%. On the contrary, cost (4.5%) and time (3.4%) generate the greatest disagreements among agents, becoming another two of the major sources of failure in construction projects, as stated in the literature [110].

Conclusions
The Spanish model related to the organization of enabling degrees and professional practice is a specific model that is not widespread in Europe. Certain academic degrees legally have their own sphere of competences that directly enable the practice of a profession in the construction sector (aeronautical, agronomic, civil, geological, industrial, mining, organizational, telecommunications and topographical engineers and technical engineers, as well as architects and technical architects), ensuring their presence in the building process, which is not the case in other countries. However, although this is not mandatory, these agents participate in the process in different countries of the EU, although under different names, but with similar roles and responsibilities (safety managers instead of health and safety planners or project and work health and safety coordinators, quantity surveyors instead of work supervisors, etc.).
The IPM contract allows technicians to focus on the functions explicitly included in their professional powers, (technical design of the projects and/or subsequent supervision or coordination of the works). This frees them from other management tasks (requirements analysis, quality plans, budgets, certifications, obtaining administrative licenses, selection of suppliers, comparison of offers, etc.), which are performed implicitly in the traditional contract, often without remuneration or specific preparation to do so.
According to the objectives outlined in this research and answering the research questions (defining and determining the relationships of power between building agents and their position in relation to the overall objectives of the project), it was necessary to select a method that allows analysis of both the qualitative and quantitative variables that make up the different scenarios, their characteristics, interdependencies between them and their limits. MACTOR is a simple but useful tool that leads to a better understanding of the power play and relationships between building agents. The added value provided by MACTOR derives mainly from the position of the agents in relation to their goals, as well as from the tactics for possible alliances and conflicts and strategic recommendations to prevent potential future scenarios based on their agreements and convergences and taking into account their disagreements and divergences. In this sense, thanks to the compilation of data from the matrices of dependences and direct influences (MDI) and valued positions (2MAO), the necessary information is obtained to understand the relationships between agents and their alignment with the objectives, which allows their behaviour to be predicted, allowing possible problems between agents, as well as deviations from the project's objectives, to be prevented.
These analyses are applicable in situations that are difficult to foresee, in which multiple actors are involved and varying interests, perspectives, and options collide. The MACTOR results explain that the influence and interdependence between active stakeholders and show how powerful the competitiveness of stakeholders is in determining the overall objectives to be formulated. The higher the competitiveness, the stronger the influence of building agents in accommodating the interests of each agent. The methodology employed is not only an instrument that helps in decision-making itself, but it also provides an instrument for group use.
The study of building agents shows the importance of project managers, as drivers (and leaders) of the projects, exerting decisive influence over the other agents. It is also worth noting the role played by public administrations (for their ability to demand solutions that comply with legality), by construction managers (as assistant project managers during the construction phase of the building), by developers (as direct clients) and by owners (as holders of the property). The designers are in intermediate positions, despite possessing the intellectual rights of the object of construction. In addition, those responsible for implementing contracts and projects accordingly and to good construction standards (project and work supervisors and control entities) are also at an intermediate level of hierarchy. On the other hand, those agents involved in implementation (main and specific contracts), are at the lower end of the scale, presenting a strong dependence on the other agents involved.
The study of the project objectives shows that correct scope is essential for good project completion. However, the other two main constraints (time and cost) are the objectives that cause the greatest conflict of interest among building agents. With regard to the agreements and disagreements that these cause between agents, the greatest convergences occur between project managers and construction managers, exceeding those that occur between them and agents with specific professional power. Conversely, the greatest divergences are between project managers and developers, owners, financial institutions and insurance companies, although these divergences are five times less than the convergences outlined. The presence of managerial agents thanks to the IPM contract allows developers to delegate operational decisions, technicians to focus on their technical goals without assuming functions not explicitly included in their contracts and professionals to be controlled.
In short, the stakeholder analysis reflects that, in an IPM contractual context, the influence of technical agents (project designers and project and work supervisors) is reduced by limiting their functions to those marked by the regulatory framework, as well as reducing the room for manoeuvre of professionals, subjected to systematised monitoring and supervision, whose mission is masked in other forms of contract (for technicians in the traditional model and for professionals in the turnkey model). In addition, prospective analysis highlights the importance of defining scope (and quality) from the early stages of a project, as well as the need to reach multilateral agreements based on the other two main project constraints: time and cost, based on the convergences and agreements that the technical performance provides. However, further research is required to adjust policies, strategies and routines that ensure potential future scenarios. Given that this study is the first study, it is recommended to continue this line of research with another tool from the prospective analysis, such as MICMAC, that is a structural prospective analysis used to study indirect relationships and MULTIPOL, which is another prospective analysis used to compare different actions or solutions for problems according to multiple criteria and policies.

Conflicts of Interest:
The authors declare no conflict of interest.