Institutional Environment, Institutional Arrangements, and Risk Identification and Allocation in Public–Private Partnerships: A Multilevel Model Analysis Based on Data from 31 Provinces in China

Abstract

As an important part of market-based reforms, the issue of “risk” has always been a part of the public–private partnership (PPP) debate, and the way in which risks are managed determines the sustainability of market-based reforms. This study systematically examines how the institutional environment and institutional arrangements affect the effectiveness of PPP risk identification and allocation. The study aims to establish a multi-source database of all publicized PPP projects in 31 provinces during the period of 2017–2021 in China, and it incorporates different levels of influencing factors, such as the institutional environment and institutional arrangements, into a framework. Through the application of a multilevel model, the impact of the political–institutional structure at the provincial level and institutional arrangements at the social capital and project levels on PPP risk identification and allocation is judged hierarchically. It is found that only social capital at the institutional environment level can directly and positively contribute to the effective identification and allocation of PPP risks, while both the degree of potential market competition and the degree of integration at the project level have a positive effect on its risk profile. In addition, there are cross-level moderating effects, with social capital, government transparency, and government–business relations positively improving risk allocation and identification through the degree of market competition, with no significant coordinating effect on the degree of contractual integration. This study goes beyond the existing one-dimensional risk profile analysis and risk typology to dismantle the “black box” of risk identification and allocation, which is a process of continuous negotiation, and to provide a sustainable governance mechanism for the market-oriented reform of public services in the institutional arena.

1. Introduction

With administrative reforms since the 1970s, public–private partnerships (PPPs) have gradually become a new tool for public service delivery that balances efficiency and quality. In PPPs, public actors and private contractors work together on projects such as infrastructure development and the delivery of public services while sharing the associated risks, costs, and benefits [1,2]. The involvement of the private sector has introduced enormous challenges and new risks to public governance [3]. Determining how to provide the sustainable development governance mechanism of PPPs has become an important issue. The long-term and complex nature of large-scale projects and contracts greatly increases the ambiguity and asymmetry of information, thus raising the risk of PPPs [4,5,6,7,8]. In practice, a proliferation of PPP failures because of improper risk identification and allocation has been reported worldwide [9,10,11]. Thus, risk and its transfer are foundational elements of PPPs [12], and they are crucial for sustainable governance. The identification, allocation, and management of risks greatly affect the success or failure of a PPP [8,13,14,15,16,17,18]. The aim of this paper is to analyze the factors, especially the institutional factors, affecting PPP risk identification and allocation by unraveling their inherent processes. Addressing these issues will contribute to the exploration of sustainable governance mechanisms for PPP projects.

Contemporary research on the risks of PPPs includes three aspects: First, a number of studies have found that it is important for both governmental and private sectors to manage risks from a project life-cycle perspective, identifying and assessing risks as early as possible, as well as allocating them to the parties that are best able to control them [13,19,20]. Second, a large number of studies have focused on the identification of categories of risk from different perspectives [21,22]. Third, several scholars have explored the factors affecting risk allocation and identification in specific theoretical frameworks [23,24,25].

In summary, while studies have noted the importance of identifying and allocating risks in PPPs, they have mainly emphasized “risk lists and menus” and “best practices” in risk allocation. However, the formulation of best practices and whether risks are indeed effectively identified and allocated in practice are completely different issues. “Effective risk allocation in PPP projects is therefore no easy job” [25]. Contemporary research either views risk as an integral part of the PPP contractual arrangement or “hides” the risk allocation in other dimensions, such as “contract renegotiation” [26] and “termination” [27,28]. Few studies have provided an in-depth discussion on the process of risk identification and allocation or the factors that influence its effective allocation. Meanwhile, contemporary studies have noted the role played by project-level institutional arrangements and the institutional environment in PPP risk [26]; they tend to view institutional arrangements as elements of risk. The impact of cross-level institutional factors on the identification and allocation of PPP risks has not been well discussed. Based on these issues, this study addresses the following questions: What are the factors that affect the effective identification and allocation of risks in PPPs? In particular, what is the impact of cross-level institutional environments and arrangements on the identification and allocation of PPP risks?

To address these questions, this study constructed a theoretical framework for understanding PPP risk governance by introducing an institutional analysis framework and formulating several theoretical hypotheses. To test assumptions, this study constructed a multi-source database of 31 provinces in China for all publicized PPP projects during the period of 2017–2021. A multi-layer linear model was applied to analyze the factors affecting the effective identification and allocation of risks in PPPs. Through this analysis, it was found that the institutional arrangement at the project level has a direct effect on the identification and allocation of PPP risks, and social capital at the institutional level directly affects the effective identification and allocation of PPP risks. Moreover, social capital, government–business relationships, and government transparency have a positive moderating effect on the degree of competition and the effective identification and allocation of risks. This study expands on the research of PPP risk governance, as well as the relationship between institutions and PPPs.

This study intends to make theoretical contributions regarding sustainable governance arrangements and mechanisms for the market-oriented reform of public services. First, in the field of PPP risk, going beyond the existing one-dimensional risk profile analysis and risk typology, this study regards PPP risk identification and allocation as a continuous negotiation process between the public and private sectors, providing a process foundation for PPP risk governance. Second, by clearly quantitatively measuring and testing the impact of institutional environment and arrangements on project risk, this study establishes theoretical and practical guidelines for constructing a sustainable PPP project governance structure.

This paper is organized as follows: First, a brief review of the relevant literature on PPP and its risk management is presented, including a discussion on the knowledge gaps. Subsequently, a theoretical framework is constructed based on PPP risk governance and institutional analysis, and relevant theoretical hypotheses are formulated. In Section 4, the data, measurements, and econometric model are presented. Section 5 reports the results of the empirical research. Finally, the relevant theoretical findings are briefly discussed.

2. Literature Review

2.1. Identification and Allocation of Risk in Public–Private Partnerships

Risk is an essential attribute of PPPs. As risk becomes the “cornerstone” of PPPs, determining how to manage risk is the key to reducing it [29,30]. Risk management can be divided into four phases: risk identification, an analysis of impacts, the response to minimize risks, and the allocation of appropriate contingencies [9,31,32]. Of these, risk identification and allocation are the most important aspects [33]. Contemporary research mainly perceives risk identification and allocation as specialized technical and managerial issues [34,35].

Risk identification is the first step in risk management and forms the basis for risk analysis and evaluation [33]. The key to risk identification is the construction of effective techniques and specialized tools for identifying risks [36]. Using different tools, various scholars have constructed and categorized a series of risk lists or catalogs [13,23,34,37,38]. For example, Grimsey and Lewis identified eight types of risk and broadly categorized them into global and elemental risks [39]. Guo et al. supplemented an analysis of mining risk with economic risk, political risk, contractor risk, and civilian risk as the core of social risk [40]. Ng and Loosemore categorized PPP risks into general risks and project risks based on the source of the risk [41]. From a game theory perspective, Ni Anna Ya categorized PPP risks into investment environment risks, project risks, and partnership risks [42]. Ke et al. identified 14 country-level risks, 7 market-level risks, and 16 project-level risks using desktop research [10]. Le et al. conducted a systematic literature review, identified 86 unique risks, and grouped them according to the project cycle [43].

Risk-sharing is the core feature that distinguishes a PPP from other traditional outsourcing projects [8]. Based on different perspectives, such as game theory [44], principal–agent theory [45], fuzzy synthetic allocation models [36], and transaction cost economics [25,46], contemporary research has developed different models for optimal risk allocation. There is a general consensus that optimal risk allocation needs to follow several simple rules, i.e., risk needs to be allocated to the party that is the most cognizant of that risk and has the strongest ability to assess it and handle it [16,24,41,42,47,48]. These principles suggest that project-level risks, such as financial risks, design and construction risks, and operational risks, should be allocated to the private sector [49]; environment-level risks, such as regulatory risks, macroeconomic risks, and political risks, should be allocated to the public sector [20,42,50]. Other risks, such as social risks, environmental risks, and force majeure, should be shared [51,52].

However, the principle and the practice of risk allocation are separate issues. The influencing factors of specific risk allocation mechanisms and arrangements have also been discussed in the literature. Albalate et al. (p. 495) argued that “countries with higher institutional quality and stability are able to engage in PPPs with fewer guarantees or less need for sharing the risks associated with demand, cost overrun, and maintenance and operation [47]. Past failures might offset this stability however”. Among these, both institutional stability and the regulation environment affect the strategy of risk allocation [53].

2.2. Effective Identification and Allocation of Risk as a Decision-Making Process

Overall, contemporary research has extensively discussed the typologies, causes, and identification and allocation of PPP risks. However, on the one hand, the effective identification and allocation of risks are very difficult. Contemporary research tends to perceive PPP risk as “endogenous”, i.e., as originating from the institutional arrangement itself, thus considering it to be an intrinsic attribute of the governance arrangement of the PPP [25,54,55]. However, the identification of risk sources and influencing factors does not help policymakers solve the decision-making problems associated with risk identification and allocation. On the other hand, current research focuses on “professional analysis and appropriate allocation” as a technical issue and has developed a number of underlying principles [39]. However, “optimal decision-making” is not always practiced. In other words, the question of whether “best practice” can be followed is itself a problem. Specific risk allocation is an outcome of the decision-making process. In practice, risks are often misallocated to parties that do not have the resources, information, or capacity to manage risks [41,56].

Moreover, although the current research has discussed the sources of PPP risks and the factors affecting the identification and allocation of risks, factors are split at the macro- and micro-levels. Any arrangement is embedded in a specific institutional environment, and the elements between the macro- and micro-levels are not compartmentalized. Therefore, contemporary research lacks an integrative framework for analyzing PPP risk identification and allocation. As Tallaki and Bracci noted, the way in which the institutional environment affects the identification, transfer, and allocation of PPP risks, particularly in developing countries, still requires further research [23].

In short, the questions of what factors and which mechanisms influence the identification and allocation of risk in PPPs have not been adequately answered. As Jin and Doloi (p. 719) stated, “risk allocation decision making was better interpreted rather than seen as a black box or only capability-driven” [25]. Opening this black box reveals that the allocation of risk is an ongoing process of consultation and negotiation. Effective and rational risk allocation is an empirical issue rather than a functional or value issue. The central question of this study, therefore, is as follows: beyond risk typology and the study of “optimal decision-making” in risk allocation, what factors affect the effective identification and allocation of PPP risks within the PPP field?

3. Institutional Environments, Institutional Arrangements, and Risk Identification and Allocation: A Theoretical Framework

3.1. Risk Governance from an Institutional Analysis Perspective

As noted earlier, risk allocation is not an ideal matching or calculation process; rather, it is a decision-making process. Such decision-making unfolds within a particular governance structure. In terms of the PPP risk management process, the risk allocation decision includes both the identification and assessment of risk, as well as the negotiation and consultation of risk allocation [33,48,57]. This process ultimately determines whether risks can be allocated effectively. Currently, more scholars have recognized that the PPP itself is a governance system with complex relationships and contracts; it is a process of shared decision-making and collaborative governance across organizations and sectors [4,23,31,37,58,59,60,61,62,63,64,65]. This means that effective risk identification and allocation requires effective governance systems and capabilities [31]. Failed PPPs often stem from failures in decision-, responsibility-, and risk-sharing mechanisms [66]. The introduction of the governance perspective also fits contemporary research, which emphasizes the role that institutions play in PPPs. A growing number of studies perceive the governance of PPPs as an institutional field, emphasizing the impact of the enabling field and institutional maturity on the success of PPPs [67,68,69,70].

By integrating these foundational perspectives, this study constructs an institutional analytical framework for analyzing PPP risk identification and allocation. Under this framework, on the one hand, the nature of PPPs is understood as an institutional arrangement for the governance of long-term and complex public affairs, of which joint decision-making and risk-sharing are inherent attributes [71]. On the other hand, as a contractual arrangement, environmental uncertainty and behavioral uncertainty are core sources of PPP risk [57,72]. Among them, diverse and effective governance mechanisms are key to achieving the effective identification and allocation of risk [23,57,73,74]. In general, the integrative framework can unravel the impact of a PPP’s institutional arrangements and the environment on the identification and allocation of PPP risk (see Figure 1).

In this framework, risk identification and allocation are a shared decision-making process between the public and private sectors. Whether risks can be identified or allocated relies on four main governance mechanisms: experience and capacity, information and communication, trust, and power dynamics. Together, the existence and functioning of these governance mechanisms form the basis for the effective identification and allocation of PPP risks. In terms of experience and capacity, in volatile and uncertain PPP projects, the project experiences, risk identification and allocation experiences, and shared governance experiences of both the public and private sectors can affect the identification and allocation of risks. In terms of information and communication, risk itself is a social construct and dialogue process. Assessing the sources and distribution of risk and identifying the party with the greater risk-taking and coping capacities require accurate and detailed information and diverse mechanisms for the sharing and communication of information [5,19]. However, conflicts of interest, institutional logic, and differing objectives between the public and private sectors cause both sectors to conceal information, leading to information asymmetry [57,64,75,76]. Effective information transfer and flow improve transparency in both sectors, especially in the government sector, and they increase the accuracy of risk identification and assessment, as well as effective risk allocation [5,57]. In terms of trust mechanisms, the biggest challenge facing long-term contracts lies in the uncertainty of the future; long-term commitment is an intrinsic attribute of PPPs [1,4]. Trust means that trusting the other party will prevent opportunistic behavior, even if opportunities rise [77]. Trust can resolve the tension between a rigid contract and managerial flexibility by structuring a relational contract that achieves contractual adaptability [24,31,58]. With the establishment of a trust mechanism, both parties are willing to share the risks in the PPP when uncertainty emerges in the future [57,77]. Finally, in terms of power dynamics, in PPPs, there is an asymmetry in resources, capabilities, and legitimacy between the public and private sectors, indicating unequal relative bargaining power, which substantially affects the negotiation and decision-making process for risk identification and allocation [41]. In general, the public sector tends to hold the advantage of relative bargaining power, making it more likely that risk is inappropriately allocated to the private sector [78]. However, the public sector can also be asymmetrically “locked in” by their suppliers, causing the private sector to shift risks to the public sector [79]. Therefore, equal power mechanisms are important for the identification and allocation of risk in PPPs [48].

As the black box of governance mechanisms for PPP risk identification and allocation has been opened, the mechanisms regarding how various levels of institutions affect the identification and allocation of PPP risks have been identified. For a long time, although a large number of studies have discussed the impact of governance models and institutional environments on PPPs, these studies did not identify the cross-level effects of the institution. The present study focuses on the impact of two levels of institutions on the identification and allocation of PPP risks: the institutional arrangement at the project level and the institutional environment level. In particular, regarding the institutional arrangement at the project level, the degree of market competition and the degree of contractual integration of PPPs are the focus of the present study. Regarding the institutional environment level, this study focuses on both formal and informal institutions, i.e., power and legal structures, as well as social capital. Next, this study provides further analyses of how these institutional elements affect both the identification and allocation of PPP risks.

3.2. Institutional Arrangements and Risk Identification and Allocation

A large body of work has analyzed the impact of institutional arrangements at the PPP project level on PPP risks and performance [80,81,82,83]. From a transaction cost economics perspective, the long-term nature of PPPs and the complexity and uncertainty of the projects may lead to large specific asset investments and incomplete contracts [25], thus increasing the level of risk exposure of each party [84,85,86]. Contemporary research has shown that the approach, complexity, type, and duration of the contract can all affect the survival of a PPP [37,86,87]. However, as noted earlier, the exploration of risk sources is not the same as decisions on risk identification and allocation. The way in which project-level institutional arrangements affect the effective identification and allocation of PPP risks needs to be discussed in depth. By focusing on project-level institutional arrangements, this study focuses on two variables: the degree of competitiveness of a PPP and the degree of contractual integration.

In terms of PPP competitiveness, sufficient competition plays an important role in the identification and allocation of risk [20,88,89]. Competition in PPPs improves the accuracy and visibility of information [74,86]. Typically, in a more competitive tendering approach, bidding among multiple private contractors makes cost information publicly available [90,91]). Moreover, the contract itself is a mechanism for communicating information. Contracts provide a formal language, a common and negotiated framework of preferences, and a common space for action, which, in turn, reduce the risk of misinterpretation between the public and private sectors, thus reducing the cost of information processing [92]. The higher the degree of competitiveness, the more channels are available for information dissemination and the more effective the information indicators will be. Further, as the level of sound competition increases, the power mechanisms between the governmental and private sectors become more equal, thus increasing the transparency of the project and reducing the incentives and ability of both parties to transfer their risks to others [74]. As a result, the high participation of qualified bidders leads to full interaction between the parties in the bidding process and “merit-based” cooperation by the governmental sector. This can effectively contribute to the efficient identification, assessment, and allocation of risks.

From the internal viewpoint of the project contract, the full life cycle of a PPP involves a number of stages, including project identification, preparation, procurement, execution, and handover; effective contract management is an important condition for the success of PPPs [89,93,94,95,96,97]. Each stage of a PPP has a different agent and responsible authorities, each of which has its own operating mechanism and mode. Therefore, the transition between phases is a complex governance process. There are also considerable differences in the risks of PPPs at different stages. On the one hand, the leading sector of PPP may be ambiguous. Unclear functional division introduces the overlapping of departmental powers, leading to the disorder of the project operation process and communication. On the other hand, during the whole process, from project design, investment, and financing to construction, operation, and maintenance, both the public and private parties need to provide input. Examples of these materials are implementation programs, output descriptions, and value-for-money evaluation reports, which are often assessed and reviewed repeatedly in multiple stages [98]. Coupled with the redundant operational processes, the uncertainty and complexity of the procedures can hinder the effective identification and allocation of project risks. In summary, the long-term and effective integration of various phases of PPPs and the clarification of management responsibilities [37], as well as the clear classification of the responsibilities of each party, can improve information symmetry between public and private sectors; they can also diversify communication channels and promote the trust level between both sectors, which, in turn, will enhance the effectiveness of the identification and allocation of risks.

Based on this, this study proposes the following hypotheses on the direct effect of institutional arrangement and risk identification and allocation at the project level:

H1: 

Institutional arrangements affect risk identification and allocation status;

H1a: 

The higher the degree of market competition, the better the risk can be identified and allocated;

H1b: 

The higher the degree of integration, the better the risk can be identified and allocated.

3.3. Institutional Environments and Risk Identification and Allocation

With the introduction of institutions into PPP research, many studies discussed the impact of institutions on the process and outcomes of PPPs [94,96,97]. The institutional environment is a key source of PPP risk [72]. The government’s ability to construct an institutional environment that is conducive to PPPs and guides contract design and implementation can substantially affect the level of risk, efficiency, and effectiveness of PPPs [7,26,53,77]. For example, from the perspective of an enabling institutional field, the ability of government departments to build effective institutional models and tools specific to the governance of PPPs is critical to the success of PPPs [68,69]. Further, the governance and decision-making process of PPP risks is embedded in a specific institutional environment. Institutional environments, such as institutional stability and the regulatory environment, affect the effective identification and allocation of PPP risks by influencing diverse PPP risk governance mechanisms [53,99]. This study focuses on two aspects of the institutional environment: the political–institutional structure and social capital.

The political–institutional structure involves two main aspects: the power structure and the property rights structure. In terms of the power structure, the negotiation process between the public and private sectors is embedded in the power and resource structures of both parties. The construction and selection of PPPs are inherently highly political, especially in China’s local governments [100,101]. For example, in China, certain PPP projects may be terminated early because of top-down political pressure [28]. The asymmetry of the power structure and political risks are the core risk sources of PPPs [57,78,98]. In terms of risk identification and allocation, the power structure configures the distribution of power and responsibility between organizational and individual actors. A clear power structure, especially the checks and balances of power, promotes citizen participation, increases government transparency, and reduces corruption [102,103]; it also promotes competition among the private sector, cuts the cost of communicating information, and raises the level of trust and co-operation, thereby improving the effectiveness of risk identification and allocation [104].

In terms of property rights, the main reasons for adopting a PPP are its high asset specificity, sizable initial costs, and substantial scale economies [105]. However, because of factors such as leadership turnover and policy changes, the original commitments may be difficult to fulfill, leading to a failure in realizing the aspired gains for the private sector. Contemporary research has shown that different legal traditions affect the choice of contract termination and expiration [27,106]. Similarly, the establishment of a legal system in the whole process of the PPP can reduce competition and behavioral uncertainty for each subject, thus reducing institutional costs. Legal frameworks and other forms of institutional support are essential to create a favorable environment for PPPs [26].

In China, a “good” political–institutional structure includes government transparency, government–business relations, and a fair judicial system. On the one hand, transparency facilitates third-party follow-up and the scrutiny of government expectations and private sector performance, as well as public oversight; thus, information asymmetries between the government and the private sector are reduced, and the effectiveness of risk allocation and identification is increased [39,107]. Transparency links democratic accountability and market efficiency, symbolizing responsible governance and public access to information. China’s transparency is generally measured by openness in decision-making, management services, implementation and results, policy interpretation and response to concerns, and disclosure by application. A higher transparency helps private organizations understand cooperation and consultation processes, facilitating risk identification and allocation.

On the other hand, China’s government–enterprise relationship is characterized by strong governmental corporatism. A cordial and clean government–business relationship is a necessary path towards improving communication efficiency and building cooperation and trust [108,109]. The Cleanliness Index measures the integrity or corruption level of government agencies. It focuses on factors such as corruption, bribery, and public trust. Under a close government–business relationship, the interaction between the government sector and the private sector can increase mutual trust and enhance private sector engagement; this can transform into support and commitment on both sides, thereby improving the effectiveness of risk identification and allocation [5,33,110]. Finally, consistent with the role of power structures, judicial arrangements are a key influencing factor of PPP risk. A fairer judiciary can increase the asymmetry of power between the public and private sectors, reduce the risk of “capture” and rent-seeking behavior, and increase the level of trust, which, in turn, enhances the effectiveness of risk identification and allocation [111].

Finally, the current research on PPPs has shown that informal institutions can “fill in” the gaps in formal property rights institutions [86]. Informal elements of the institutional environment, i.e., social capital, play an important role in PPP risk identification and allocation. Social capital refers to the social links that promote collective behavior, including the three dimensions of trust, norms, and networks [112]. Social capital can influence PPP risk identification and allocation through diverse mechanisms. First, social capital implies the participation of diverse social actors, which can promote the participation and competition between private and public sectors in PPPs and reduce the level of information asymmetry [74,95]. Second, trust is an endogenous element of social capital. Social capital can increase the level of general trust between the government and the private sector, which, in turn, improves the effectiveness of the identification and allocation of risk in PPPs [68,77]. Finally, participation is also a process of supervision and inhibiting the effects of power mechanisms. With increasing social capital, the relative bargaining power between the government and the private sector changes, which, in turn, improves the effectiveness of PPP risk allocation and identification.

Based on this, this study proposes the following hypotheses regarding the direct effect of the institutional environment on risk identification and allocation:

H2: 

The institutional environment affects the effectiveness of risk identification and allocation;

H2a: 

The higher the social capital, the better the risk can be identified and allocated;

H2b: 

The more transparent the government sector, the better the risk can be identified and allocated;

H2c: 

The higher the level of government care, the better the risk can be identified and allocated;

H2d: 

The higher the degree of government cleanliness, the better the risk can be identified and allocated;

H2e: 

The higher the degree of judicial impartiality, the better the risk can be identified and allocated.

3.4. Institutional Arrangements, Institutional Environment, and Risk Identification and Allocation

From an institutional analysis perspective, institutions are hierarchical; lower-level institutions are embedded in higher-level institutions. Similarly, in the governance structure of PPPs, the mechanism is embedded in the institutional arrangement at the project level, and the project-level institutional arrangement is embedded in the institutional environment [20]. For a long time, from an institutional analysis perspective, numerous studies analyzed the impact of the institutional environment and institutional arrangements on PPPs. However, the embeddedness of institutions has been neglected.

Following this logic, there are several explorations of contemporary research. On the one hand, given the institutional diversity in PPP governance, mutually reinforcing effects exist between different institutions [67]. A positive alignment of institutional elements such as legal origin, current institutions, and institutional stability can substantially contribute to the success of a PPP project [113]. Moreover, different combinations of institutional elements will construct unique pathways leading to mature PPP market performance [67]. For example, contract renegotiation is closely related to risk misallocation [114]. Soecipto and Verhoest used the QCA methodology to analyze how combinations of institutional elements at macro-, meso-, and micro-levels can avoid contract renegotiation [26]. They found that a single institutional condition does not ensure contract stability, but a combination of several institutional conditions does. On the other hand, to ensure contractual stability, PPPs are usually expected to have a favorable business environment and a secure financing system; however, in reality, these conditions rarely exist at the same time. Interestingly, even in the absence of certain conditions, a combination of different institutional elements can still contribute to contract stability. Thus, different levels of institutional factors may be nested, and they interact with each other. For example, Quelin et al. showed that the experience and capacity of the public sector have a complementary effect with institutional quality on the private sector’s decision-making of the PPP involvement level [115]. However, Wang et al. found that the business environment of the local government plays a moderating role between government characteristics (e.g., fiscal pressure) and the level of private investment in PPPs [116].

These explorations suggest that the institutional environment establishes “meta-governance” for PPPs, which forms the basis for promoting the institutional enabling field and institutional maturity [68]. The institutional environment may be a moderating variable between institutional arrangements and the effective identification and allocation of PPP risks. In other words, in a high-quality institutional environment with great transparency, close government–business relationships, a fair judicial system, and a high level of social capital, the effect of project-level institutional arrangements—including the degree of competition and the contractual integration—on the effectiveness of PPP risk identification and allocation can be enhanced. However, once the governance arrangements for PPPs become constrained by institutional voids [85], the impact of the competitiveness and contractual integration at the project level on effective risk identification and allocation is reduced.

Based on this, this study formulates the following hypotheses on the moderating effect of the institutional environment:

H3: 

The institutional environment positively enhances the effect of institutional arrangements on the effectiveness of the identification and allocation of risk;

H3a: 

The political–institutional environment positively moderates the relationship between the degree of market competition and the effectiveness of the identification and allocation of risk;

H3b: 

Social capital positively moderates the relationship between the degree of market competition and the effectiveness of the identification and allocation of risk;

H3c: 

The political–institutional environment positively moderates the relationship between the degree of project integration and the effectiveness of the identification and allocation of risk;

H3d: 

Social capital positively moderates the relationship between the degree of project integration and the effectiveness of the identification and allocation of risk.

4. Methodology

4.1. Data Context

In the 30 years following their introduction to China in the 1980s, PPPs experienced a rise and fall [117], with sustainable development becoming a goal. China introduced a series of policies to promote PPPs in order to alleviate local fiscal pressures and finance infrastructure, which triggered the phenomenon of PPP “overheating”. In fact, many PPP projects in the Chinese market have failed because of unsuccessful risk avoidance, such as the financing failure of a wastewater treatment plant in Jiangsu Province and the Shanghai Dafang water plant that ended in government buyback [118]. That is, the “landing rate” of PPPs is not as expected by the local government, and the way in which risk can be avoided has always been a key concern. In 2017, documents from the Ministry of Finance, the Development and Reform Commission, the State-Owned Assets Supervision and Administration Commission, one line and three chambers, and other major national economic authorities were released intensively. This marks the turning year of PPPs from overheated development to rational development, and PPP development is equipped with institutionalized constraints and rationalized development. By the end of November 2021, the cumulative number of projects in the project management library of the national PPP comprehensive information platform had reached 10,209, with a total project investment of CNY 16.1 trillion. However, projects are always accompanied by risks. The strategy that China has chosen to deal with risks lies in measuring the risk level through VFM data and using it as a core indicator of whether a PPP can be executed. More critically, China’s PPP data are unique in that the measurement and assessment of risk emphasizes a full life-cycle process. That is to say, the entire process of the project, from the pre-project stage, construction stage, and operation stage to the handover of the project, is integrated, and a comprehensive indicator is generated as the fundamental source of data. The final result of risk measurement is the feedback of the whole process, from risk identification to practice, which effectively reflects the characterization of risk as a decision-making process.

4.2. Data Collection

The data of this study mainly originate from the PPP projects disclosed by the China Public Private Partnerships Center (CPPPC). A database covering PPPs in 31 provinces has been established, with a time distribution from 2017 to 2021. The reason for choosing the start year of 2017 is that, since 2017, the Ministry of Finance and other major national economic authorities have intensively issued documents to regulate local government PPPs. The five-year period from 2017 to 2021 forms an important stage for PPPs to shift from high-speed growth to high-quality development. In addition, despite the widespread use of PPPs, in many regions of the world, PPP data are not disclosed to the public. However, in China, PPP data are publicly available and provide substantial advantages. First, the scale and scope of PPPs in China have been widely used in PPP-related research, providing excellent data reliability [22,117]. Second, the database has complete, scientific, and clear measurement standards and indicators, providing a solid foundation for research. Third, the central government requires that all local government PPP projects be publicly entered into a database and that the data be complete. The database of PPPs establishes the basis for project-level institutional arrangements and the measurement of the results of risk identification and allocation.

In the database, the basic information of each PPP project is provided by local financial departments in all provinces and cities nationwide to CPPPC. All PPP projects entered into the CPPPC system are included in our database. First, regarding the indicators, the Ministry of Finance organizes industry experts to conduct further evaluations of submitted PPP projects based on VFM [119,120], focusing on whether the risks of the projects are well identified and allocated. Second, we categorize each PPP project using Python technology, encompassing not only the risk identification and allocation indicators specific to each project but also collecting indicators such as the degree of potential competition, the degree of full life-cycle integration, the procurement method, the industry that it belongs to, the province, and the year corresponding to the project level, generating a preliminary catalog. Third, provincial data collection is relatively complex. The number of social organizations was selected from the National Bureau of Statistics to measure the social capital of each province. Regarding formal institutions, relevant reports from national think tanks (China Government-Business Relationship Report (2018) and Chinese Government Transparency Report) were synthesized, and relevant indicators were selected. Finally, the pieces of basic information for each PPP project were matched one by one, and these variables collectively formed the overall structure of the database. After the above comprehensive matching of various types of project information and the processing of missing values, a total of 2939 PPP projects from 2017 to 2021 were selected as the sample, covering 31 provinces.

4.3. Measurement

(1) 

Dependent Variables

Whether risks are adequately identified and reasonably allocated between both the government and social capital is a key issue in the whole life cycle of PPP operation. Each PPP examined in this study is assigned an independent rating of the effectiveness of risk identification and allocation (RISK) using data from the China VFM database.

The CPPPC is an information disclosure platform, and we collect data from this platform using Python and identify the risk identification and allocation of each PPP. The measure is scored by the local level finance department (or PPP center) together with the industry authority where the particular PPP project is located, providing an independent and objective after-the-fact evaluation. It should be emphasized that this indicator is the score and judgment of a single PPP project. This index is not directly related to the real risk level of the project, and it is only related to whether the risk of the project itself is clearly identified and effectively distributed to both the public and private sectors. The score ranges from 0 to 100. The measure is further classified into five grades, i.e., favorable, more favorable, average, more unfavorable, and unfavorable, and the corresponding scores range from 0 to 20, 21 to 40, 41 to 60, 61 to 80, and 81 to 100 points, respectively, reflecting the effectiveness of the risk identification and allocation of a particular PPP.

(2) 

Independent Variables

The independent variables at the project level are mainly the degree of potential competition (COMP) and the degree of life-cycle integration (INTE). The data source for the measurement of both variables is the PPP database, and the measurements are shown in

Table 1. Variables and measurements.

Variable		Variable Name	Measurement	Data Source
Dependent Variable		Risk Identification and Allocation	Specific values The larger the value, the better the degree of risk identification and allocation.	China Public Private Partnerships Center (CPPPC)
Independent Variables	Program Level—Institutional Arrangements	Degree of Potential Competition	Specific values The larger the value, the better the degree of potential competition	China Public Private Partnerships Center (CPPPC)
		Degree of Full Life-Cycle Integration	Specific values The larger the value, the better the degree of full life-cycle integration	China Public Private Partnerships Center (CPPPC)
	Institutional Environment Level	Number of Registered Social Organizations	Specific values The larger the value, the better the degree of social capital	National Bureau of Statistics of China (NBS)
		Government Care—Government Care Index (2018)	Specific values The larger the value, the better the government–enterprise relationship	China Government-Business Relations Report (2018)
		Government Clean—Cleanliness Index (2018)	Specific values The lower the value, the higher the government’s cleanliness	China Government-Business Relations Report (2018)
		Policy Transparency (2018)	Specific values The larger the value, the higher the policy transparency	Government Transparency Index Report
		Judicial Impartiality (2018)	Specific values The larger the value, the higher the degree of judicial impartiality	China Judicial Civility Index Report
Control Variables		Procurement Methods	1: Invitation to tender 2: Single-source procurement 3: Competitive negotiation 4: Competitive consultation 5: Open tendering	China Public Private Partnerships Center (CPPPC)
		Industry	1: Infrastructure 2: Ecological conservation 3: Social and livelihood 4: Other	China Public Private Partnerships Center (CPPPC)
		Modus Operandi	1: BOT 2: TOT 3: ROT 4: OM 5: BOO 6: Other	China Public Private Partnerships Center (CPPPC)
		Return Mechanism	1: Feasibility gap grant 2: Government expense 3: User fees	China Public Private Partnerships Center (CPPPC)
		PPP Stage	1: The preparatory phase 2: The procurement phase 3: The implementation phase	China Public Private Partnerships Center (CPPPC)

Source: authors’ own production.

. Provincial independent variables are measured at two levels: social capital and the institutional environment. To measure social capital (SOCP) at the provincial level, scholars usually use social participation, social norms, social trust, and social reciprocity. The number of registered social organizations can reflect the development of social capital in a region to a certain extent. Therefore, this study adopts the number of registered social organizations to measure social capital. Regarding the formal institutional environment, the Government Care Index (CARE) and the Government Integrity Index (GOIN) are chosen as measures of the “cordiality and cleanness” of government–business relations. Among them, the Government Care Index is measured by the government’s concern for enterprises, the government’s services to enterprises in infrastructure, financial services, market intermediaries, e-government efficiency, and the tax burden of enterprises. The Government Integrity Index is measured by the agency price of the food safety license and the Baidu corruption index. The indices are obtained from the annual research report on the ranking of China’s urban political and business relations published by the Research Center for Government-Enterprise Relations and Industrial Development of the National Development and Strategy Institute of Renmin University of China. This is widely used by local governments to improve the business environment and better match the market.

Moreover, the variables of policy transparency (TRAS) and judicial impartiality (JUIM) are measured based on the report of the Index of Transparency and Judicial Impartiality of the Chinese Government. Transparency involves examining the pre-disclosure of major decision-making and the openness of normative documents at all government levels, the disclosure of management services, the execution and disclosure of the results, the indicators for policy interpretation and responsiveness to public concerns, and the accessibility of disclosure channels upon request and the standardization of responses. The Policy Transparency Index stems from a 2018 report jointly produced by the National Research Center of Rule of Law Index at the Chinese Academy of Social Sciences and the Innovation Project Team of Rule of Law Index from the Law Institute. This report extensively investigates and evaluates the level of openness in government affairs across various administrative tiers.

Additionally, the score ratio of the subjective and objective indicators of judicial impartiality data is 9:1. The objective indicator data are obtained from the 2018 annual work report issued by the Higher People’s Court and the People’s Procuratorate to the provincial people’s Congress, and the subjective data are obtained from questionnaire interviews examining judicial power, judicial culture, judicial openness, parties’ rights, civil judicial procedures, legal professionalization, criminal judicial procedures, administrative judicial procedures, judicial corruption containment, and evidence systems. The “China Justice Index” is a quantitative assessment tool for the rule of law developed by the Judicial Civilization Collaborative Innovation Center, which is part of the national “2011 Plan” and the “Double First-Class” Construction Plan. The development of this index reflects the satisfaction of the people with the state of local judicial civilization, providing a basis for strengthening the construction of judicial civilization throughout the country.

(3) 

Control Variables

According to the previous literature, PPP performance is also related to factors such as the industrial sector, procurement method, modus operandi, return mechanism, current stage of the PPP project, and duration of the project construction period. In this study, these are controlled for and included in a model. In addition, to control for the effects of different years, this study also controls for the time factor by including the year in the model. The industrial sector is divided into the following four sectors: infrastructure, ecological conservation domain, social and livelihood, and other. The procurement methods are divided into the following five categories: invitation to tender, single-source procurement, competitive negotiation, competitive consultation, and open tender. The modus operandi is divided into the following six categories: BOT, TOT, ROT, OM, BOO, and other. The return mechanism is divided into the following three categories: feasibility gap grant, government expense, and user fees. The stage of the PPP project is divided into the following three phases: the preparatory phase, the procurement phase, and the implementation phase. These factors are included in the model through dummy variables, with “other” as the reference group for the industrial sector, “open tender” as the procurement method, “other” as the modus operandi, “user fees” as the return mechanism, “implementation phase” as the stage of the PPP project, and 2017 as the reference year.

At this point, the variable data of different stages (see Table 1) are collected, and a multi-source database is established.

4.4. Model Specification and Centering of Explanatory Variables

The variables used in this study involve two levels: the PPP project level (level-1), such as RISK, COMP, and INTE, and the provincial level (level-2), such as the SOCP, CARE, GOIN, TRAS, and JUIM. Level-1 is nested within level-2, and, for this nested data structure, the error term will violate the independence and homoskedasticity assumptions if ordinary multiple regression analyses are used. Therefore, this study uses a multilevel analysis to test the research hypotheses.

Specifically, this study employs five models in sequence to stepwise conduct tests of the research hypotheses presented above [121]: “M1” (null model), where the calculation of the intra-group correlation coefficient (ICC) is used to determine whether a multilevel analysis is needed; “M2” (random coefficient model), which is used to analyze the direct effect of the level-1 explanatory variables on the dependent variable (RISK); “M3” (model with intercept as the dependent variable), which is used to analyze the direct effect of the level-2 explanatory variables on RISK; “M4” (context model), which is used to analyze the direct effect of both the level-1 and level-2 explanatory variables on RISK; and “M5” (full model), which is used to test both the first- and second-level explanatory variables on RISK and simultaneously test for cross-level interactions between level-1 and level-2. Models (1), (2), (3), (4), and (5) are as follows:

(1)

M1

$$\begin{array}{l}\mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l}\text{-}1:\\ \mathit{R}\mathit{I}\mathit{S}\mathit{K}={\mathit{\beta }}_{0\mathit{j}}+{\mathit{r}}_{\mathit{i}\mathit{j}}\\ \mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l}\text{-}2:\\ {\mathit{\beta }}_{0\mathit{j}}={\mathit{\gamma }}_{00}+{\mathit{u}}_{0\mathit{j}}\end{array}$$

(2)

M2

$$\begin{array}{l}\mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l}\text{-}1:\\ \mathit{R}\mathit{I}\mathit{S}\mathit{K}={\mathit{\beta }}_{0\mathit{j}}+{\mathit{\beta }}_{1\mathit{j}}\mathit{C}\mathit{O}\mathit{M}{\mathit{P}}_{\mathit{i}\mathit{j}}+{\mathit{\beta }}_{2\mathit{j}}\mathit{I}\mathit{N}\mathit{T}{\mathit{E}}_{\mathit{i}\mathit{j}}+{\mathit{\beta }}_{3\mathit{j}}\mathit{P}\mathit{E}\mathit{R}{\mathit{I}}_{\mathit{i}\mathit{j}}+{\mathit{\beta }}_{.\mathit{j}}{\mathit{Z}}_{\mathit{i}\mathit{j}}+{\mathit{r}}_{\mathit{i}\mathit{j}}\\ \mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l}\text{-}2:\\ {\mathit{\beta }}_{0\mathit{j}}={\mathit{\gamma }}_{00}+{\mathit{u}}_{0\mathit{j}}\\ {\mathit{\beta }}_{1\mathit{j}}={\mathit{\gamma }}_{10}+{\mathit{u}}_{1\mathit{j}}\\ {\mathit{\beta }}_{2\mathit{j}}={\mathit{\gamma }}_{20}+{\mathit{u}}_{2\mathit{j}}\\ {\mathit{\beta }}_{3\mathit{j}}={\mathit{\gamma }}_{30}+{\mathit{u}}_{3\mathit{j}}\\ {\mathit{\beta }}_{.\mathit{j}}={\mathit{\gamma }}_{.0}\end{array}$$

(3)

M3

$$\begin{array}{l}\mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l}\text{-}1:\\ \mathit{R}\mathit{I}\mathit{S}\mathit{K}={\mathit{\beta }}_{0\mathit{j}}+{\mathit{r}}_{\mathit{i}\mathit{j}}\\ \mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l}\text{-}2:\\ {\mathit{\beta }}_{0\mathit{j}}={\mathit{\gamma }}_{00}+{\mathit{\gamma }}_{01}\mathit{S}\mathit{O}\mathit{C}{\mathit{P}}_{\mathit{j}}+{\mathit{\gamma }}_{02}\mathit{C}\mathit{A}\mathit{R}{\mathit{E}}_{\mathit{j}}+{\mathit{\gamma }}_{03}\mathit{G}\mathit{O}\mathit{I}{\mathit{N}}_{\mathit{j}}+{\mathit{\gamma }}_{04}\mathit{T}\mathit{R}\mathit{A}{\mathit{S}}_{\mathit{j}}+{\mathit{\gamma }}_{05}\mathit{J}\mathit{U}\mathit{I}{\mathit{M}}_{\mathit{j}}+{\mathit{u}}_{0\mathit{j}}\end{array}$$

(4)

M4

$$\begin{array}{l}\mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l}\text{-}1:\\ \mathit{R}\mathit{I}\mathit{S}\mathit{K}={\mathit{\beta }}_{0\mathit{j}}+{\mathit{\beta }}_{1\mathit{j}}\mathit{C}\mathit{O}\mathit{M}{\mathit{P}}_{\mathit{i}\mathit{j}}+{\mathit{\beta }}_{2\mathit{j}}\mathit{I}\mathit{N}\mathit{T}{\mathit{E}}_{\mathit{i}\mathit{j}}+{\mathit{\beta }}_{3\mathit{j}}\mathit{P}\mathit{E}\mathit{R}{\mathit{I}}_{\mathit{i}\mathit{j}}+{\mathit{\beta }}_{.\mathit{j}}{\mathit{Z}}_{\mathit{i}\mathit{j}}+{\mathit{r}}_{\mathit{i}\mathit{j}}\\ \mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l}\text{-}2:\\ {\mathit{\beta }}_{0\mathit{j}}={\mathit{\gamma }}_{00}+{\mathit{\gamma }}_{01}\mathit{S}\mathit{O}\mathit{C}{\mathit{P}}_{\mathit{j}}+{\mathit{\gamma }}_{02}\mathit{C}\mathit{A}\mathit{R}{\mathit{E}}_{\mathit{j}}+{\mathit{\gamma }}_{03}\mathit{G}\mathit{O}\mathit{I}{\mathit{N}}_{\mathit{j}}+{\mathit{\gamma }}_{04}\mathit{T}\mathit{R}\mathit{A}{\mathit{S}}_{\mathit{j}}+{\mathit{\gamma }}_{05}\mathit{J}\mathit{U}\mathit{I}{\mathit{M}}_{\mathit{j}}+{\mathit{u}}_{0\mathit{j}}\\ {\mathit{\beta }}_{1\mathit{j}}={\mathit{\gamma }}_{10}+{\mathit{u}}_{1\mathit{j}}\\ {\mathit{\beta }}_{2\mathit{j}}={\mathit{\gamma }}_{20}+{\mathit{u}}_{2\mathit{j}}\\ {\mathit{\beta }}_{3\mathit{j}}={\mathit{\gamma }}_{30}+{\mathit{u}}_{3\mathit{j}}\\ {\mathit{\beta }}_{.\mathit{j}}={\mathit{\gamma }}_{.0}\end{array}$$

(5)

M5

$$\begin{array}{l}\mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l}\text{-}1:\\ \mathit{R}\mathit{I}\mathit{S}\mathit{K}={\mathit{\beta }}_{0\mathit{j}}+{\mathit{\beta }}_{1\mathit{j}}\mathit{C}\mathit{O}\mathit{M}{\mathit{P}}_{\mathit{i}\mathit{j}}+{\mathit{\beta }}_{2\mathit{j}}\mathit{I}\mathit{N}\mathit{T}{\mathit{E}}_{\mathit{i}\mathit{j}}+{\mathit{\beta }}_{3\mathit{j}}\mathit{P}\mathit{E}\mathit{R}{\mathit{I}}_{\mathit{i}\mathit{j}}+{\mathit{\beta }}_{.\mathit{j}}{\mathit{Z}}_{\mathit{i}\mathit{j}}+{\mathit{r}}_{\mathit{i}\mathit{j}}\\ \mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l}\text{-}2:\\ {\mathit{\beta }}_{0\mathit{j}}={\mathit{\gamma }}_{00}+{\mathit{\gamma }}_{01}\mathit{S}\mathit{O}\mathit{C}{\mathit{P}}_{\mathit{j}}+{\mathit{\gamma }}_{02}\mathit{C}\mathit{A}\mathit{R}{\mathit{E}}_{\mathit{j}}+{\mathit{\gamma }}_{03}\mathit{G}\mathit{O}\mathit{I}{\mathit{N}}_{\mathit{j}}+{\mathit{\gamma }}_{04}\mathit{T}\mathit{R}\mathit{A}{\mathit{S}}_{\mathit{j}}+{\mathit{\gamma }}_{05}\mathit{J}\mathit{U}\mathit{I}{\mathit{M}}_{\mathit{j}}+{\mathit{u}}_{0\mathit{j}}\\ {\mathit{\beta }}_{1\mathit{j}}={\mathit{\gamma }}_{10}+{\mathit{\gamma }}_{11}\mathit{S}\mathit{O}\mathit{C}{\mathit{P}}_{\mathit{j}}+{\mathit{\gamma }}_{12}\mathit{C}\mathit{A}\mathit{R}{\mathit{E}}_{\mathit{j}}+{\mathit{\gamma }}_{13}\mathit{G}\mathit{O}\mathit{I}{\mathit{N}}_{\mathit{j}}+{\mathit{\gamma }}_{14}\mathit{T}\mathit{R}\mathit{A}{\mathit{S}}_{\mathit{j}}+{\mathit{\gamma }}_{15}\mathit{J}\mathit{U}\mathit{I}{\mathit{M}}_{\mathit{j}}+{\mathit{u}}_{1\mathit{j}}\\ {\mathit{\beta }}_{2\mathit{j}}={\mathit{\gamma }}_{20}+{\mathit{\gamma }}_{21}\mathit{S}\mathit{O}\mathit{C}{\mathit{P}}_{\mathit{j}}+{\mathit{\gamma }}_{22}\mathit{C}\mathit{A}\mathit{R}{\mathit{E}}_{\mathit{j}}+{\mathit{\gamma }}_{23}\mathit{G}\mathit{O}\mathit{I}{\mathit{N}}_{\mathit{j}}+{\mathit{\gamma }}_{24}\mathit{T}\mathit{R}\mathit{A}{\mathit{S}}_{\mathit{j}}+{\mathit{\gamma }}_{25}\mathit{J}\mathit{U}\mathit{I}{\mathit{M}}_{\mathit{j}}+{\mathit{u}}_{2\mathit{j}}\\ {\mathit{\beta }}_{3\mathit{j}}={\mathit{\gamma }}_{30}+{\mathit{u}}_{3\mathit{j}}\\ {\mathit{\beta }}_{.\mathit{j}}={\mathit{\gamma }}_{.0}\end{array}$$

One of the properties of a multilevel analysis is the ability to estimate the random effects of the explanatory variables on the impact of the explained variables at level-1; this includes the two parameters of intercept and slope. The random effect of the intercept reflects the average difference in the PPP project risk across provinces; the random effect of the slope reflects the provincial differences in the magnitude of the effect of the explanatory variables on the explained variables. For the intercept to be representative of the average value of RISK across provinces, the explanatory variables at level-1 must be centered appropriately [122,123].

In the multilevel model with two levels, the explanatory variables at level-1 are subjected to two forms of centering: group-mean centering and grand-mean centering. In grand-mean centering, the sample mean is subtracted from the predictor score of each PPP project (i.e., $X_{ij}-\overline{X}$). In group-mean centering, the predictor mean for the province to which the PPP project belongs is subtracted from the predictor scores for each PPP project within that province (i.e., $X_{ij}-\overline{X_j}$). However, these two forms of centering do not carry the same importance for intercept adjustment: while group-mean centering allows for the intercept to be the mean value of the PPP project risk across provinces, grand-mean centering is subject to provincial differences in the explanatory variables, thus making the intercept the adjusted means of the PPP project risk.

In a multilevel analysis, if the purpose of the research is to explore cross-level interactions, the explanatory variables at level-1 should be group-mean-centered so that the intercepts represent the differences in the explained variable across groups. In this case, to counteract the effect of group differences in the explanatory variables caused by group-mean centering on the explained variable, it is necessary to place the group means of the explanatory variables at level-1 into a regression equation using the intercept as the outcome [124]. In contrast, explanatory variables at level-2 are generally grand-mean-centered, which allows for the intercept of the regression equation to effectively reflect the means of the explained variable [122]. This centering approach not only reduces the multicollinearity problem between variables [125] but also allows for the intercept terms of the models to be closer to the actual RISK scores, which makes it easier to interpret the results of the data analysis.

Because M2, M3, M4, and M5 used in this study contain explanatory variables at different levels, the analysis follows the centering strategy described above, whereby the explanatory variables in each model are centered: the explanatory variables at level-2 are grand-mean-centered, and the explanatory variables at level-1 are group-mean-centered. At the same time, in the analysis of M3, M4, and M5, the explanatory variables at level-1 are aggregated and put into the intercept estimation equation to adjust for the estimation bias caused by the use of group-mean centering.

Statistical analyses were conducted using HLM 6.0.8 software [126].

5. Results

5.1. Results of Descriptive Statistics

Table 2. Summary statistics and data sources.

	Mean	Mean	SD	Data Source
RISK	Risk identification and allocation	80.625	6.684	CPPPC a
Institutional Arrangements
COMP	Degree of potential competition	80.961	6.877	CPPPC
INTE	Degree of full life-cycle integration	82.761	6.299	CPPPC
Institutional environment
SOCP	Number of registered social organizations (logarithmically processed)	10.135	0.600	NBSC b
CARE	Government Care Index	18.849	7.296	CGBRR c (2018)
GOIN	Cleanliness Index	10.284	6.303	CGBRR d (2018)
TRAS	Policy Transparency Index	70.359	7.470	CGBRR e (2018)
JUIM	Judicial impartiality	69.297	1.339	CGTR f (2018)
Control variables
PREI	Duration of cooperation	21.340	6.633	CPPPC
DCNS	Dummy variables for infrastructure	0.635	-	CPPPC
DECL	Dummy variables for the ecological conservation domain	0.225	-	CPPPC
DSEC	Dummy variables for social and livelihood domains	0.127	-	CPPPC
DITT	Dummy variables for invitation to tender	0.002	-	CPPPC
DSSP	Dummy variables for single-source procurement	0.007	-	CPPPC
DCNE	Dummy variables for competitive negotiation	0.002	-	CPPPC
DCCN	Dummy variables for competitive consultation	0.069	-	CPPPC
DBOT	Dummy variable for BOT	0.776	-	CPPPC
DTOT	Dummy variable for TOT	0.064	-	CPPPC
DROT	Dummy variable for ROT	0.065	-	CPPPC
DOOM	Dummy variable for OM	0.002	-	CPPPC
DBOO	Dummy variable for BOO	0.005	-	CPPPC
DQBZ	Dummy variables for feasibility gap grant	0.738	-	CPPPC
DZFF	Dummy variables for government expense	0.226	-	CPPPC
DPRE	Dummy variables for preparatory phase	0.076	-	CPPPC
DPRC	Dummy variables for procurement phase	0.411	-	CPPPC
D018	Dummy variables for the year 2018	0.195	-	CPPPC
D019	Dummy variables for the year 2019	0.208	-	CPPPC
D020	Dummy variables for the year 2020	0.263	-	CPPPC
D021	Dummy variables for the year 2021	0.044	-	CPPPC

Notes: ^a China Public Private Partnerships Center. ^b National Bureau of Statistics of China. ^c China Government-Business Relations Report. ^d China Government-Business Relations Report. ^e Government Transparency Index Report. ^f China Judicial Civility Index Report.

reports the mean and standard deviation of each variable, and

Table 3. Descriptive statistics of variables and correlation coefficients.

	1 a	2 b	3 b	4 b	5 b	6 b	7 a	8 a
1. RISK a	-
2. SOCP b	0.136 **	-
3. CARE b	0.141 **	0.062	-
4. GOIN b	0.039 *	−0.099	−0.299	-
5. TRAS b	0.050 **	0.152	0.443 *	−0.473 **	-
6. JUIM b	0.024	0.151	0.441 *	−0.231	0.358 *	-
7. COMP a	0.664 **	0.109 **	0.136 **	0.050 **	0.063 **	0.010	-
8. INTE a	0.742 **	0.102 **	0.134 **	0.026	0.071 **	−0.016	0.612 **	-
9. PERI a	−0.039 *	−0.043 *	0.066 **	−0.049 **	−0.037 *	−0.148 **	0.008	0.035

Notes: ^a Variables at level-1. ^b Variables at level-2. * p < 0.05. ** p < 0.01.

reports the correlation coefficients between the bivariate variables. It should be noted that the correlations between the variables at level-2 were calculated using data from 31 provinces; the cross-level correlation coefficients were calculated by disaggregating the data from the provincial variables to individual PPP projects in terms of the PPP projects (n = 2926). When the sample size of a study is large, the power of the statistical test increases. This means that if there is a real effect in the study, a large sample size makes it more likely that the effect will be detected. Therefore, for the test of correlation coefficients, the p-value significance is likely to be below common significance levels (such as 0.05) due to the large sample size, making the test results easily judged as statistically significant. However, this does not necessarily mean that the detected effect is important in practical terms because the statistical significance might simply be due to the large sample size. Therefore, the significance levels of the various correlation coefficients for the explanatory variables at level-1 in Table 3 should be considered for reference purposes only.

The data presented in Table 3 show that, among the five explanatory variables at level-2 (i.e., SOCP, CARE, GOIN, TRAS, and JUIM), only CARE and TRAS (r = 0.443), CARE and JUIM (r = 0.441), TRAS and GOIN (r = −0.473), and TRAS and JUIM (r = 0.358) are significantly correlated with each other.

Among the four variables at level-1, significant correlations exist between RISK and COMP (r = 0.664), INTE (r = 0.742), and PERI (r = 0.039), and the directions of the correlations are as expected. Moreover, there is a significant positive correlation between COMP and INTE (r = 0.612), and there is no significant correlation between PERI and COMP or INTE.

Regarding the correlation between cross-level variables, there are significant positive correlations between RISK and SOCP (r = 0.136), CARE (r = 0.141), GOIN (r = 0.039), and TRAS (r = 0.050), while there is a positive but non-significant correlation with JUIM. Except for the non-significant positive correlation between INTE and GOIN, significant positive correlations exist between the two explanatory variables at level-1, namely, INTE and COMP, and the five explanatory variables at level-2, namely, SOCP, CARE, GOIN, TRAS, and JUIM. Except for the significant positive correlation with CARE, PERI has a significant negative correlation with the remaining four explanatory variables at level-2.

In the multicollinearity tests, the variance inflation factor values were calculated for all explanatory variables. All variance inflation factor values remained well below the acceptable threshold of 5.0 (ranging from 1.24 to 1.62).

5.2. Results of Multilevel Modeling Analysis

(1) 

M1 (Null Model): Differences in RISK Across Provinces

The null model, a multilevel model with no predictor, was performed, allowing for an assessment of whether there was a significant between-province variation in RISK. In a multilevel model, the variance of the dependent variable can be decomposed into within-group variance and between-group variance, and only when the between-group variance component reaches significance can subsequent cross-level analyses be conducted. The null model is modeled as model (1) in Section 4.

The analytical results presented in

Table 4. Results of the multilevel analysis of RISK.

	M1	M2	M3	M4	M5
Fixed effect a
$\mathbf{For} \mathbf{INTRCPT}\mathbf{1} {\mathit{\beta }}_{\mathbf{0}}$
$\mathbf{INTRCPT}\mathbf{2}\mathbf{,} {\mathit{\gamma }}_{\mathbf{00}}$	$\begin{array}{c}{80.803}^{ ***}\\ \left(158.008\right)\end{array}$	$\begin{array}{c}{79.884}^{ ***}\\ \left(86.280\right)\end{array}$	$\begin{array}{c}{80.811}^{ ***}\\ \left(169.285\right)\end{array}$	$\begin{array}{c}79.881{ }^{***}\\ \left(88.394\right)\end{array}$	$\begin{array}{c}{79.955}^{ ***}\\ \left(83.834\right)\end{array}$
$\mathbf{TORG}\mathbf{,} {\mathit{\gamma }}_{\mathbf{01}}$			$\begin{array}{c}0.032{ }^{†}\\ \left(1.831\right)\end{array}$	$\begin{array}{c}{0.034}^{ *}\\ \left(2.054\right)\end{array}$	$\begin{array}{c}{0.040}^{ *}\\ \left(2.317\right)\end{array}$
$\mathbf{CARE}\mathbf{,} {\mathit{\gamma }}_{\mathbf{02}}$			$\begin{array}{c}0.030\\ \left(0.739\right)\end{array}$	$\begin{array}{c}0.038\\ \left(0.970\right)\end{array}$	$\begin{array}{c}0.031\\ \left(0.582\right)\end{array}$
$\mathbf{FAIR}\mathbf{,} {\mathit{\gamma }}_{\mathbf{03}}$			$\begin{array}{c}0.008\\ \left(0.148\right)\end{array}$	$\begin{array}{c}0.014\\ \left(0.246\right)\end{array}$	$\begin{array}{c}0.016\\ \left(0.240\right)\end{array}$
$\mathbf{TRAS}\mathbf{,} {\mathit{\gamma }}_{\mathbf{04}}$			$\begin{array}{c}0.077\\ \left(1.096\right)\end{array}$	$\begin{array}{c}0.080\\ \left(1.097\right)\end{array}$	$\begin{array}{c}0.096\\ \left(1.239\right)\end{array}$
$\mathbf{CLEN}\mathbf{,} {\mathit{\gamma }}_{\mathbf{05}}$			$\begin{array}{c}0.488\\ \left(1.131\right)\end{array}$	$\begin{array}{c}0.436\\ \left(1.062\right)\end{array}$	$\begin{array}{c}0.477\\ \left(1.069\right)\end{array}$
$\mathbf{For} \mathbf{COMP} \mathbf{slope} {\mathit{\beta }}_{\mathbf{1}}$
$\mathbf{INTRCPT}2\mathbf{,} {\mathit{\gamma }}_{\mathbf{10}}$		$\begin{array}{c}{0.311}^{ ***}\\ \left(8.803\right)\end{array}$		$\begin{array}{c}{0.311}^{ ***}\\ \left(8.790\right)\end{array}$	$\begin{array}{c}{0.320}^{ ***}\\ \left(8.361\right)\end{array}$
$\mathbf{TORG}\mathbf{,} {\mathit{\gamma }}_{\mathbf{11}}$					$\begin{array}{c}{0.004}^{ *}\\ \left(1.987\right)\end{array}$
$\mathbf{CARE}\mathbf{,} {\mathit{\gamma }}_{\mathbf{12}}$					$\begin{array}{c}0.008{ }^{*}\\ \left(2.131\right)\end{array}$
$\mathbf{FAIR}\mathbf{,} {\mathit{\gamma }}_{\mathbf{13}}$					$\begin{array}{c}-0.033\\ \left(-1.277\right)\end{array}$
$\mathbf{TRAS}\mathbf{,} {\mathit{\gamma }}_{\mathbf{14}}$					$\begin{array}{c}{0.010}^{ *}\\ \left(2.022\right)\end{array}$
$\mathbf{CLEN}\mathbf{,} {\mathit{\gamma }}_{\mathbf{15}}$					$\begin{array}{c}0.004\\ \left(0.790\right)\end{array}$
$\mathbf{For} \mathbf{COMB} \mathbf{slope} {\mathit{\beta }}_{\mathbf{2}}$
$\mathbf{INTRCPT}2\mathbf{,} {\mathit{\gamma }}_{\mathbf{20}}$		$\begin{array}{c}0.537{ }^{***}\\ \left(17.990\right)\end{array}$		$\begin{array}{c}0.539{ }^{***}\\ \left(18.012\right)\end{array}$	$\begin{array}{c}0.520{ }^{***}\\ \left(14.952\right)\end{array}$
$\mathbf{TORG}\mathbf{,} {\mathit{\gamma }}_{\mathbf{21}}$					$\begin{array}{c}0.001\\ \left(0.582\right)\end{array}$
$\mathbf{CARE}\mathbf{,} {\mathit{\gamma }}_{\mathbf{22}}$					$\begin{array}{c}-0.005\\ \left(-1.002\right)\end{array}$
$\mathbf{FAIR}\mathbf{,} {\mathit{\gamma }}_{\mathbf{23}}$					$\begin{array}{c}0.001\\ \left(0.061\right)\end{array}$
$\mathbf{TRAS}\mathbf{,} {\mathit{\gamma }}_{\mathbf{24}}$					$\begin{array}{c}0.003\\ \left(0.687\right)\end{array}$
$\mathbf{CLEN}\mathbf{,} {\mathit{\gamma }}_{\mathbf{25}}$					$\begin{array}{c}0.003\\ \left(0.686\right)\end{array}$
$\mathbf{For} \mathbf{PERI} \mathbf{slope} \mathbf{(}{\mathit{\gamma }}_{\mathbf{30}}$)		$\begin{array}{c}-0.042{ }^{*}\\ \left(-2.753\right)\end{array}$		$\begin{array}{c}-0.042{ }^{*}\\ \left(-2.723\right)\end{array}$	$\begin{array}{c}-0.050{ }^{**}\\ \left(-3.060\right)\end{array}$
$\mathbf{For} \mathbf{Z} \mathbf{slope} \mathbf{(}{\mathit{\gamma }}_{\mathbf{.0}}$)	No	Yes	No	Yes	Yes
Random effect b
	Variance Component
$\mathbf{INTRCPT}1\mathbf{,} {\mathit{\tau }}_{\mathbf{00}}$	$\begin{array}{c}{7.532}^{ ***}\\ \left(624.804\right)\end{array}$	$\begin{array}{c}8.047{ }^{***}\\ \left(1556.679\right)\end{array}$	$\begin{array}{c}{7.932}^{ ***}\\ \left(602.598\right)\end{array}$	$\begin{array}{c}8.232{ }^{***}\\ \left(1461.867\right)\end{array}$	$\begin{array}{c}8.256{ }^{***}\\ \left(1502.291\right)\end{array}$
$\mathbf{COMP} \mathbf{slope}\mathbf{,} {\mathit{\tau }}_{\mathbf{11}}$		$\begin{array}{c}{0.029}^{ ***}\\ \left(157.371\right)\end{array}$		$\begin{array}{c}{0.029}^{ ***}\\ \left(157.296\right)\end{array}$	$\begin{array}{c}{0.026}^{ ***}\\ \left(137.809\right)\end{array}$
$\mathbf{COMB} \mathbf{slope}\mathbf{,} {\mathit{\tau }}_{\mathbf{22}}$		$\begin{array}{c}0.0146{ }^{***}\\ \left(95.948\right)\end{array}$		$\begin{array}{c}{0.014}^{ ***}\\ \left(95.951\right)\end{array}$	$\begin{array}{c}{0.015}^{ ***}\\ \left(79.432\right)\end{array}$
$\mathbf{PERI} \mathbf{slope}\mathbf{,} {\mathit{\tau }}_{\mathbf{33}}$		$\begin{array}{c}0.002{ }^{***}\\ \left(59.735\right)\end{array}$		$\begin{array}{c}0.002{ }^{***}\\ \left(59.739\right)\end{array}$	$\begin{array}{c}{0.003}^{ ***}\\ \left(59.865\right)\end{array}$
$\mathbf{level}\text{-}\mathbf{1}\mathbf{,} {\mathit{\sigma }}^{\mathbf{2}}$	37.156	14.946	37.154	14.943	14.921
Deviance	18,961.086	16,404.006	18,975.430	16,414.925	16,492.292
Number of estimated parameters	2	11	2	11	11

Notes: ^a t-values are reported in parentheses. ^b ${\chi }^2$-values are reported in parentheses. † p < 0.1; * p < 0.05; ** p < 0.01; *** p < 0.001.

show that the between-province variation in RISK (τ₀₀ = 7.532) reaches significance (χ² = 624.804, df = 30, p < 0.001), indicating that RISK is significantly different across provinces. In addition, the within-province variance is σ² = 37.156. The above between- and within-province variances reach ICC = 0.168, indicating that around 17% of the variation in RISK could be attributable to differences between provinces. According to Hox et al. (2017) [127], when ICC > 0.058, the between-group variance is not negligible, i.e., there is a significant difference in the mean values of RISK between different provinces. In other words, when analyzing RISK, it is necessary to consider the differences and characteristics of the provinces.

(2) 

M2 (Stochastic Coefficient Model): the Relationship between Hierarchical Variables and Project Risk

Once the need for multilevel modeling is established, the hypotheses of this study are tested in turn. The first test is whether the two explanatory variables at level-1 (COMP and INTE) affect RISK. Model (2) is shown in Section 4.

For the brevity of presentation, Z is used to represent the dummy variables for the industrial sector, procurement method, modus operandi, return mechanism, stage of the PPP project, and year (see Table 2 for details). In the model,Z and PERI are control variables. γ10 and γ20 represent the slopes of the impact of the two explanatory variables on RISK; if the parameter estimates reach the significance level, the two explanatory variables have a significant influence. Table 4 shows that COMP and INTE reach significance (γ₁₀ = 0.311, p < 0.001; γ₂₀ = 0.537, p < 0.001). This indicates that the more competitive the project market and the better the project integration management, the more apparent the risk points of the project and the clearer the risk allocation. This shows that research hypotheses H1, H1a, and H1b are supported.

In terms of random effects, the results of τ₀₀, τ₁₁, and τ₂₂ show that the estimated value of the intercept term variance, τ₀₀, in this model is 7.532 (χ² = 624.804, p < 0.001), indicating the presence of a significant intercept term variance component. In addition, the slope term variance components of both COMP and INTE also reach significance (p < 0.001). Because different intercept terms exist across provinces, it can be hypothesized that the explanatory variables at the provincial level may have a direct effect on RISK.

(3) 

M3 (Intercept Model): Relationship between Provincial Level Variables and RISK

The intercept model, i.e., the model with the mean of level-1 as the dependent variable, assumes that there are no explanatory variables at level-1 and instead predicts RISK based on the explanatory variables at level-2 (SOCP, CARE, GOIN, TRAS, and JUIM). Model (3) is shown in Section 4.

As shown in Table 4, SOCP, CARE, GOIN, TRAS, and JUIM have a positive effect on RISK; however, except for the coefficient of SOCP, which is significant (p = 0.040), the coefficients of the other four variables are not significant. Therefore, it can be assumed that the explanatory variables at the provincial level have no direct effect on RISK, except for SOCP. In other words, research hypotheses H2b, H2c, H2d, and H2e are not supported, while H2 and H2a are supported. Moreover, in terms of random effects, the estimate of the intercept term variance τ₀₀ remains statistically significant (χ² = 602.598, p < 0.001). This result indicates that there are other factors that affect RISK at the provincial level that are not considered in this study. Future research can explore the factors that may affect RISK at the provincial level in more depth.

(4) 

M4 (Context Model): the Relationship between Different Levels of Variables and RISK

The two aforementioned models estimate the effects of the explanatory variables on RISK at level-1 and level-2 but do not answer the question as to what the direct effects of each variable are if variables at different levels are considered at the same time. The context model can test the direct effects of different levels of explanatory variables when they are entered into the model at the same time. Model (4) is shown in Section 4.

As shown in Table 4, when considering the direct effects of the explanatory variables at two different levels simultaneously, the effects of SOCP, CARE, GOIN, TRAS, and JUIM at the provincial level are quite similar to the results of the aforementioned intercept model: except for the coefficient of SOCP, which is significant (p = 0.040), the coefficients of the other four variables of the regression coefficients are not significant. This result indicates that the variables related to the provincial government still do not have a direct impact on RISK when the variables at level-1 are considered at the same time, i.e., hypotheses H2b, H2c, H2d, and H2e of this study are not supported.

Furthermore, both COMP and INTE at level-1 also reach significance (γ₁₀ = 0.311, p < 0.001; γ₂₀ = 0.539, p < 0.001). This result is the same as that of the aforementioned M2 (random coefficient model), where only level-1 explanatory variables are considered. This reconfirms that research hypotheses H1, H1a, and H1b are supported.

Taken together, the explanatory variables at level-1 have significant explanatory power when the direct effects of the explanatory variables at different levels are considered, whereas the governmental explanatory variables at level-2 have no direct explanatory power. It is worth noting that none of the four models explored so far consider the cross-level interactions between the explanatory variables at different levels. Model M5 can further test for the existence of cross-level interactions.

(5) 

M5 (Full Model): the Interaction of Two Levels of Variables

The results of the above analysis indicate that the four government-related variables at level-2 have no direct effect on RISK, but whether the explanatory variables at level-2 have a moderating effect on the relationship between the explanatory variables at level-1 and RISK remains to be tested. That is, whether the cross-level interaction between the level-2 and level-1 explanatory variables affects RISK needs to be tested. The full model can be tested for the cross-level interactions and direct effects of both levels of variables, which are modeled in model (5).

As shown in Table 4, in addition to the significant direct effect of SOCP on RISK (γ₀₁ = 0.040, p < 0.05), there is a significant cross-level interaction between SOCP and COMP (γ₁₁ = 0.004, p < 0.10); as the coefficient of the interaction term has the same positive sign as that of COMP, SOCP can positively strengthen the effect of COMP on RISK, and H3b is supported. The cross-level interaction between CARE and COMP also reaches significance (γ₁₂ = 0.008, p < 0.05), and the cross-level interaction between TRAS and COMP similarly reaches significance (γ₁₅ = 0.010, p < 0.10). Judging from the sign of the coefficients, both CARE and TRAS positively strengthen the effect of COMP on RISK. Therefore, H3 is fully supported, while H3a is only partially supported. In addition, as shown in Table 4, neither SOCP nor the four variables on the government side have a significant interaction with INTE; therefore, hypotheses H3c and H3d are not supported.

The direct effects of the two independent variables at level-1 (COMP and INTE) on RISK are not much different from M4, and both reach significance.

6. Discussion

By combining the analysis results of the various models, the following findings can be obtained: Regarding the variables of institutional arrangement at the project level, the degree of market competition and the degree of integration have a significant positive impact on the effectiveness of risk identification and allocation in PPPs (see Figure 2). This result proves the core hypothesis of this paper, namely, that the institutional arrangement at the PPP project level directly impacts the effectiveness of the identification and allocation of risk. This suggests that, on the one hand, over the life cycle of a project, different stakeholder networks are dynamically derived from various stages, from planning, design, and construction to operation and maintenance [68]. Among these, the absence of the government’s role or slow changes in its functions are common occurrences at various stages. Contractors will seek opportunism under contractual ambiguity, leading to the emergence of complex organizational, political, and governance challenges, such as cheap construction and corruption [128]. All of these challenges will introduce risk ambiguity, information asymmetry, communication difficulties, and mistrust, which, in turn, affect the effective identification and allocation of risks. The ability of actors to construct coherent governance structures can substantially affect the effectiveness of risk identification and allocation. On the other hand, a lack of competition can lead to information asymmetry between public and private parties. A high degree of competition in the market increases the frequency of interaction and trust between the public and private parties, and it fundamentally enhances the transparency of the contract formation process, for example, by disclosing more hidden cost information [129]. Thus, in the bargaining and negotiation process, both public and private parties will have a better perception of risk.

At the institutional environment level, this study simultaneously identifies the pathways through which the institutional environment influences the identification and allocation of risk and re-emphasizes the importance of social capital. Indeed, studies on PPPs have shown that risk is closely linked to trust (see Figure 2). Because of the complexity of PPPs, it is difficult for participants to foresee possible contingencies. The existence of trust allows actors to communicate information, thus avoiding abuse of power and cost-shifting, which generates confidence in PPPs. As a foundational source of social trust, this study finds that social capital strengthens participation, competition, and monitoring and promotes the balance of power between the public and private sectors; this, in turn, enhances the effectiveness of risk identification and allocation.

It is important to note that many studies have argued that formal institutional environments, such as the legal regulatory environment, institutional quality, and judicial system, play a direct and important role in PPP governance. On the one hand, a number of studies have assessed the impact of various factors on PPP risk identification and allocation without considering the interactions between different levels. Many of these studies directly attribute shortcomings to the absence of a suitable legal framework, economic and political instability, and operational issues [130,131]. Sastoque et al. contend that Colombia has yet to utilize PPP mechanisms for social infrastructure due to a lack of clear legislation. On the other hand, research has uncovered nested relationships across various levels, with Soecipto analyzing multiple combinations of factors from macro-, meso-, and micro-perspectives that determine contract renegotiation. The business environment of local governments, along with the experience and capabilities of the public sector, may serve as a mediating factor between government characteristics such as financial pressures and the level of private investment in PPPs [115,116]. However, PPP risk identification and allocation are rarely placed within a broader institutional context, leading to an inadequacy in considering comprehensive factors and their interactions. Contrasting with these findings, this study finds that these formal institutional arrangements do not directly affect the effective identification and allocation of PPP risks (see Figure 2). This is an important theoretical finding that can be further explored in the future.

Although there are certain exceptions, contemporary research focuses mainly on the singular impact of PPP project-level institutional arrangements or institutional environments on PPPs. However, the cross-level institutional analysis framework constructed in this study suggests that the governance mechanisms of PPP projects are embedded within a layered institutional system (see Figure 3). The data results indicate that the provincial-level institutional environment, i.e., social capital, government transparency, and the Government Care Index, have a positive cross-level moderating effect on the relationship between competition and effective risk identification and allocation. In an environment with high social capital, a good business environment, and high government transparency, the effects of information and power balance caused by competition are amplified, and the effective identification and allocation of risks can be better achieved. In other words, a reinforcing effect exists between high social capital and strong institutional quality and competition at the project level. This finding also provides new evidence for the complementary effect between the government’s PPP governance capacity and the institutional environment. However, it is important to note that the variables of the institutional environment at the provincial level do not have a cross-level moderating effect of project management integration on the effective identification and allocation of PPP risks. A possible explanation for this observation is that project management integration is an internal management issue of the project, and it has a relatively weak link to the external environment; thus, it is not reinforced or inhibited by the institutional environment at the provincial level.

7. Conclusions

Since the 1970s, market-based reforms have become important tools to reform public governance. How to make market-based reforms stable and sustainable has been a key concern in both academia and practice. With the involvement of the private sector, “risk” exists in almost all discussions on the performance of PPPs [6], and how risks can be effectively identified and shared is a key element in PPPs. Current research focuses on two aspects: discussing the source, level, and impact of risk and defining the types of risk, as well as the “best principles” of risk allocation, from technical and managerial perspectives. However, the answer to the question of which factors affect the effective identification and allocation of PPP risks remains unclear. This paper continues previous discussions on the correlation between the institutional environment, institutional arrangements, and institutional performance; it constructs an integrated institutional analytical framework for PPP risk identification and allocation by understanding the process of risk identification and allocation as a risk governance process.

This study innovatively integrates the institutional environment and institutional arrangement into one framework, utilizes multi-layer linear models, and systematically analyzes the factors affecting the identification and allocation of risk, thus filling a theoretical gap in the PPP field. Using PPP data from 31 provinces in China, this study finds that, in terms of project-level institutional arrangements, the higher the degree of market competition, the clearer the perception of risk and the more scientific the allocation of risk between both parties. Similarly, over the whole life cycle of a PPP, the higher the degree of integration at each stage, the more effective the identification and allocation of risks. In terms of the institutional environment, the higher the level of social capital, the more effective the identification and allocation of risk. However, while the cross-level moderating effects of social capital, government transparency, and government–business relations positively moderate the enabling effect of the degree of market competition on risk allocation and identification, they have no moderating effect on the degree of integration.

In terms of policy implications, first, to better identify and allocate risks in PPPs, the government should construct a sustainable institutional infrastructure for the marketization of public services. On the one hand, the social–institutional environment is the basic component of risk management in PPP projects. Governments should promote social credit systems to build trust, using measures such as strengthening credit information management and establishing punishment mechanisms for dishonesty. This can reduce cooperation costs in PPPs and define clear government–enterprise boundaries and rules for fair competition and regulation. Private organizations need to strengthen their own credit and communicate with the public sector to better manage project risks and ensure the sustainability of PPP projects. On the other hand, the government should not neglect the construction of a formal institutional structure. A balanced power structure and well-defined property rights are essential conditions for reducing PPP risks, and they rely on improved government–business relations and judicial system regulation. Therefore, the government can establish a positive and stable cooperative relationship with the private sector by demonstrating care and setting up communication channels with the public; the relevant laws and regulations can be presented to the public in a more open, transparent, and detailed manner, thus building a political and institutional environment that is conducive to the long-term development of PPP projects. The government should participate in formulating a clear market entry and exit mechanism, which is very important to attract capable partners and ensure the continuity of the project, allowing both parties to choose high-quality partners and reduce the risk of capacity. The private sector should implement the changes in laws and regulations in a timely manner and negotiate through institutionalized channels. At the same time, in the public–private partnership market, the professionalism and competitiveness of project services should be strengthened to form effective PPP risk control.

In addition, at the project level, the government should focus on the contract management process to save market transaction costs so as to avoid the failure of PPPs due to an unstable contract connection. This involves building a competitive market and strengthening the integration of project design, investment and financing, construction and operation, and maintenance. When necessary, the government can introduce third-party risk assessment agencies to provide professional services. More importantly, such institutional arrangements rely heavily on social capital and the construction of a cordial and clean relationship between the government and business. Therefore, governments should not build the institutional environment or institutional arrangement in isolation but, instead, should fully understand the nested relationship between the market-oriented institutional arrangement and the institutional environment. When constructing institutional environments, governments should consider market laws and development needs to avoid institutional lags or excessive intervention. Furthermore, market mechanisms can guide the continuous optimization and improvement of institutional environments. This also encourages practitioners to play their role and find an institutional environment suitable for their own project development.

Inevitably, this study has certain limitations that will provide avenues for future research. First, the quality of risk identification and allocation in this study is derived from expert ratings in the VFM database, which may result in measurement errors and affect the results. The results of risk identification and allocation and final risk transfer may be different. Even if a particular risk is reasonably allocated, it is still possible that the government or the private sector will transfer that risk to the other party during the execution of the subsequent contract. This further suggests that the allocation of risk needs to be identified and measured more clearly. Second, this study does not consider the public in risk allocation. Indeed, for both public and private sectors, “reasonable” risk allocation may ultimately come at the cost of the public bearing the ultimate risk, which is also true for an “unreasonable” risk allocation [61,132]. However, this study specifically focuses on the allocation of risk between the public and private sectors; the role of the public in the identification and allocation of risk, as well as the question of whether the public may take an inappropriate risk, needs to be further explored. Third, this study focuses on PPPs in China and the provincial institutional environment. However, PPPs in China may be unique [117]. Moreover, because of differences in the institutional regimes among different countries, their possible impact on risk identification and allocation may also be different [57]. Therefore, the generalizability of the findings needs to be further examined.