Digital Empowerment and Risk Management: Dual-Path Mechanisms and Boundary Conditions for the Sustainable Transformation of the Construction Industry

Abstract

The construction industry, a global economic pillar and carbon emission giant, faces a critical gap between digital transformation and risk management, which ultimately undermines the sector’s capacity for risk management. This study combines social technical systems theory with the technology–organization–environment framework, using panel data from Chinese listed construction firms to explore how digital transformation affects project risk management. Key findings reveal that digital transformation significantly boosts risk management through two distinct pathways. While environmental governance capacity and green innovation efficiency both serve as significant mediators, the study identifies a notable disparity in the driving forces: digital transformation exerts a stronger impact on green innovation efficiency (17.8%) compared to environmental governance (4.4%). However, the resulting mediating effects of these two paths are found to be remarkably similar (0.0060 vs. 0.0068). Furthermore, labor investment efficiency is identified as a critical boundary condition, with a threshold effect (−0.385) below which the benefits of digital transformation weaken. These findings provide empirical evidence from Chinese context regarding the “technology-institution” co-evolution mechanism in construction. While centered on China, the study offers valuable insights for global stakeholders on how to harness digitalization to mitigate project risks and enhance sustainability.

1. Introduction

Project risk management is an important part of project management. It aims to identify, assess, respond to, and monitor uncertain factors that may be encountered during project implementation through systematic methods, so as to ensure the smooth achievement of project objectives. In a rapidly changing business environment, project risk management is particularly important. It is not only related to the success or failure of the project, but also directly affects the organization’s strategic objectives and sustainable development [1].

The global construction industry is at a critical juncture, facing an increasingly complex risk landscape characterized by intensifying market volatility, stricter environmental regulations, and mounting expectations from various stakeholders [2]. Construction is a vital pillar of the global economy, contributing approximately 13% of worldwide GDP. However, its traditional development model is facing unprecedented challenges. The industry has long been criticized for its fragmented operations, multidisciplinary nature, and high complexity, and has been regarded as a laggard in technological transformation [2,3]. This situation has exposed the systemic vulnerabilities of traditional risk management approaches in construction projects [4]. Furthermore, the construction sector accounts for approximately 40% of global energy consumption and generates almost 33% of carbon emissions, making it a critical domain for addressing climate change and advancing sustainable development [1].

Meanwhile, the digital economy, centered on information technology and data, is flourishing, presenting traditional industries with a historic opportunity to transform and upgrade. The construction industry is undergoing a profound transformation driven by next-generation digital technologies such as building information modelling (BIM), the Internet of Things (IoT), big data, and artificial intelligence (AI), a process known as digital transformation. This shift is seen as essential for advancing the industry’s industrialization, digitization, and intelligent development, and will ultimately steer it toward high-quality growth [5]. By integrating scientific risk management with the opportunities presented by digital transformation, the construction industry can proactively identify potential risks, develop effective response measures, reduce the likelihood and impact of risks, and ensure the successful completion of projects according to plan, budget, and quality requirements, laying a solid foundation for sustainable development [1].

This study selects listed construction companies on China’s A-share market as empirical samples for three reasons: Firstly, according to the National Bureau of Statistics, China was the world’s largest construction market in 2023, with total output exceeding 30 trillion yuan, nearly one-third of the global total. Its transformation practices, therefore, exhibit significant scale effects and industry representativeness. Secondly, guided by its dual carbon goals, China has established the world’s most comprehensive green building policy framework. This framework provides a unique institutional setting for exploring the collaborations between digitalization and sustainable development. Thirdly, Chinese listed companies maintain transparent disclosure standards and are increasingly providing robust reporting on ESG and digital metrics [5], which offers a high-quality data foundation for empirical testing. Furthermore, Chinese construction firms are heavily involved in infrastructure development in over 150 countries along the Belt and Road routes. Their risk management expertise has considerable spillover value for global engineering practices.

However, despite the immense potential of digital technologies and sustained corporate investment in them, there remains a significant gap between technology adoption and improvements in risk management effectiveness [5]. The academic community lacks a clear understanding of the mechanisms by which digitalization enhances core performance in construction projects, particularly regarding the project risk management level. Due to the dynamic, complex, and interdisciplinary nature of construction projects, risk management remains challenging. Existing research predominantly focuses on the isolated benefits of specific digital technologies (e.g., BIM for clash detection) or on the direct impact of digital transformation on individual outcomes, such as green innovation [2]. A core theoretical question remains unresolved: Given the industry’s long-standing operational fragmentation and resistance to technological integration [6], how does digital transformation, a systemic change encompassing technology, organization, and environment, ultimately translate into enhanced project risk management capabilities in construction? Is this impact constrained by the efficiency of internal resource allocation within enterprises? To unravel this “black box”, this study integrates theoretical perspectives and deconstructs the underlying mechanisms by mapping “theory-mechanism-variable” relationships. We propose a theoretical model encompassing both “compliance-driven”, “innovation-driven” and “resource efficiency” pathways:

First, drawing on the environmental dimension of the technology–organization–environment (TOE) framework and the dynamic capabilities perspective of the resource-based view (RBV), we identify environmental governance capability as the first mediating pathway. This pathway reveals how digital transformation strengthens environmental compliance by enhancing a firm’s capacity to sense and reconfigure resources in response to environmental fluctuations.

Second, grounded in sociotechnical systems (STS) theory and the technological dimension of the TOE framework, we establish green innovation efficiency as the second mediating pathway. This pathway is based on the logic that technology must collaborate with social systems; specifically, digital transformation drives green innovation by breaking down information silos and facilitating knowledge sharing.

Finally, adopting the resource heterogeneity perspective of the RBV, we introduce labor investment efficiency as a critical moderating variable. This design aims to capture the boundary conditions imposed by the efficiency of internal human resource allocation on the effectiveness of digital transformation.

The following is the theoretical integration and research logical framework diagram of this paper (Figure 1).

This study makes several important contributions. First, in terms of theory, it combines sociotechnical systems theory, the technology–organization–environment (TOE) framework and resource-based view (RBV) [7] to reveal the transmission mechanism linking “digital capabilities–sustainable practices–risk management”, thereby filling a theoretical gap in intermediary pathway research within construction management. Secondly, in practice, it provides Chinese construction enterprises with a synergistic pathway of “technology empowerment–governance optimization–innovation enhancement”, supporting the implementation of the “Digital China” and “dual carbon” strategies. Finally, on a global scale, we distil a transferable “institutionally driven digital transformation” paradigm by leveraging China’s massive market scale and policy innovation practices. This paradigm offers an operational framework for emerging economies (such as those in Southeast Asia and Latin America) to help them avoid the transformation trap of prioritizing technology over governance, thereby enabling high-quality, sustainable engineering construction.

The structure of this paper is as follows: Section 2 elaborates on the sociotechnical systems theory, the technology–organization–environment framework, and the resource-based view. It explains how these theories provide a robust foundation for the study. Section 3 conducts a systematic literature review and proposes research hypotheses based on this review. Section 4 outlines the research methodology. It details the data sources, variable definitions, and model construction, all of which are supported by the theoretical framework. Section 5 presents the results of the empirical analysis. These include tests of the benchmark model and its impact mechanisms, as well as robustness checks. Section 6 discusses the empirical findings in light of the theoretical foundations and existing research conclusions. Section 7 concludes with recommendations and future directions.

2. Theoretical Foundations

This study draws primarily on three well-established and complementary theoretical frameworks: Social-technical systems theory, the technology–organization–environment framework, and the resource-based view. Together, these theories provide a robust foundation for understanding the multifaceted phenomenon of digital transformation and its impact on organizational capabilities and project performance.

2.1. Social-Technical Systems (STS) Theory

The STS theory emphasizes that technological (tools, processes, platforms) and social (structures, culture, human resources) subsystems must be synergistically optimized to achieve organizational effectiveness [8]. Chen et al. (2019) [7] further situate the system within an institutional environment (e.g., regulations, stakeholders, and economic conditions), elucidating how those multiple institutional forces shape technology adoption. In the context of multi-stakeholder collaboration in construction projects, this theory offers a valuable perspective on how digital technologies become embedded in social structures and influence organizational capabilities. This theoretical perspective is particularly significant because construction projects are collections of sociotechnical systems involving collaboration among numerous stakeholders within complex technological environments, and the successful implementation of digital tools (e.g., BIM, IoT) depends not only on technical specifications but also on the human and organizational context.

2.2. Technology–Organization–Environment (TOE) Framework

The TOE framework suggests that a company’s adoption and implementation of new technologies, as well as the effectiveness of these technologies, are influenced and constrained by factors across three dimensions: technology, organization, and environment. The technology dimension focuses on the technology’s own characteristics, such as compatibility, complexity, and relative advantage. The organizational dimension involves the company’s scale, structure, management processes, strategy, and culture. In contrast, the environmental dimension encompasses industry competition, government policies, regulatory frameworks, and relationships with stakeholders such as suppliers and customers. The framework’s strength lies in its systematic approach, which avoids treating technological change as a standalone event. Instead, it embeds technological transformation within broader organizational structures and external contexts for holistic examination.

The digital transformation of the construction industry aligns perfectly with the analytical dimensions of the TOE framework. Digital transformation is a comprehensive process that encompasses digital infrastructure (technological dimension), digital strategy, and digital culture (organizational dimension), and is driven by policy and market environments (environmental dimension) [9]. Similarly, enhancing project risk management involves applying technological tools (e.g., BIM for risk simulation), optimizing organizational processes (e.g., data-driven risk decision-making), and adopting agile responses to environmental risks (e.g., policy changes and market volatility). Therefore, the TOE framework provides an ideal theoretical lens through which to conceptualize digital transformation as a multidimensional construct and to examine its impact pathways systematically.

2.3. Resource-Based View (RBV)

The RBV suggests that firms are diverse collections of resources and capabilities, and that they gain a sustainable competitive advantage by possessing valuable, scarce, difficult-to-imitate, and unique resources and capabilities [10]. Resources accumulated through digital transformation, such as data analytics and BIM platforms, must be translated into tangible performance through dynamic capabilities, i.e., the ability to integrate and reconfigure resources to respond to change [11]. In the construction sector, these capabilities manifest as rapid risk identification and optimized resource allocation (Najafi et al., 2020) [12], and digital technologies can enhance them through real-time data and collaborative platforms [13]. Cultivating a digital culture and recruiting digital talent enables enterprises to develop robust, data-driven decision-making and collaborative innovation capabilities [14].

2.4. Logical Integration of Theory and Research Focus

These theories can be integrated to show that the technological capabilities enabled by digital transformation (e.g., IoT monitoring and predictive analytics) require collaboration with social systems to manage risks effectively. Therefore, building on these foundations, this study constructs an integrative theoretical model that organically integrates STS theory, TOE framework, and RBV. Adhering to the principle of “matching theoretical attributes to variable mechanisms”, we trace the theoretical origins and logical deductions for the key paths in our model.

• The mediating role of environmental governance capability: Based on the TOE environmental dimension and RBV dynamic capabilities

From the environmental dimension of the TOE framework, firms must respond to external pressures, such as “dual carbon” policies and environmental regulations. However, pressure alone is insufficient to generate performance; it requires dynamic capabilities viewed through the RBV lens as a mediator. Digital transformation extends beyond mere software acquisition; it involves leveraging digital technologies (e.g., IoT monitoring) to build a capacity for sensing, seizing, and reconfiguring environmental resources. In essence, digital transformation enhances a firm’s environmental dynamic capabilities, enabling proactive adaptation to external regulatory shocks and thereby reducing compliance risks. Consequently, environmental governance capability emerges as the inevitable outcome of technology responding to environmental pressures.

• The mediating role of green innovation efficiency: Based on STS theory and the TOE technological dimension

STS theory posits that the mere implantation of technology cannot automatically generate value; rather, it requires the co-evolution of “technical systems” and “social systems”. Through the technological lens of the TOE framework, digital transformation breaks down departmental barriers (information silos) in traditional construction. The embedding of technology must integrate with internal social structures (such as processes and collaboration mechanisms). Only by promoting cross-departmental knowledge sharing and reconfiguration can green innovation efficiency be substantially improved. Thus, this efficiency is a product of technology-organization synergy, rather than a mere byproduct of technology.

• The moderating role of labor investment efficiency: Based on RBV resource heterogeneity and the TOE organizational dimension

RBV suggests that differences in firm performance stem from resource scarcity and inimitability. In digital transformation, data serves as a core resource, yet its value extraction relies heavily on high-quality human capital. Labor investment efficiency serves as the metric for a firm’s ability to allocate this scarce human resource, constituting a moderating mechanism for the effectiveness of digital transformation. High-efficiency firms can effectively match digital assets with high-skilled labor (resource complementarity), thereby amplifying the transformation effects. In contrast, low-efficiency firms face the dilemma of being “data-rich but effectiveness-poor”.

This theoretical integration renders the study’s foundation more rigorous, in-depth, and comprehensive. It not only explains how digital transformation drives risk management through the “dual wheels” of compliance and innovation but also unveils the internal mechanism of “human resource allocation efficiency” as a moderator, thereby achieving a logical mapping from theoretical integration to variable operationalization.

3. Literature Review and Hypothesis Development

3.1. Digital Transformation in Construction and Project Risk Management Level

The construction industry has traditionally been slow to adopt technology due to fragmented supply chains and organizational inertia [15]. Against the backdrop of China’s digital economy, technologies such as BIM, the IoT, and AI are driving paradigm shifts in design, construction, and management [16]. However, their effectiveness is limited by policy environments, digital culture, and talent structures [4,5]. Digital capabilities lay the foundation for risk management by optimizing information flow, visualization, and collaboration [13]. Traditional risk management, which relies on historical data and expert experience, is unable to address novel, complex risks [4,17]. Digital technologies enable real-time monitoring and predictive analytics [18,19]; BIM reduces design conflicts [20]; and the IoT enhances on-site safety monitoring [13]. Nevertheless, the systematic enhancement of risk management capabilities through sustainable pathways within corporate governance requires further exploration.

In the construction industry, digital transformation refers to the systematic integration of information, computing, communication, and connectivity technologies to transform business processes, organizational models, and value-creation methods [21,22]. This transformation goes far beyond the mere introduction of individual technologies, encompassing the deep integration and application of a range of technologies, including BIM and digital twins, the Internet of Things (IoT), big data analytics, and artificial intelligence (AI). Project risk management capability is defined as a project team’s systematic ability to identify, analyze, address, and monitor uncertainties, ensuring that the project achieves its intended performance objectives in terms of cost, schedule, quality, and safety. We argue that digital transformation can directly and significantly improve project risk management capabilities. Digital technologies can transform risk management paradigms by breaking down information silos, providing real-time data, and enhancing collaboration. For example, BIM enables early automated detection of design conflicts, thereby preventing costly rework in the construction phase and associated safety risks. IoT sensors continuously monitor structural health, equipment status, and environmental safety, enabling proactive issuance of risk warnings. These applications shift risk management from reactive, post-event responses to proactive, pre-emptive prevention and dynamic, in-process control [23]. Therefore, digital project management platforms can facilitate a transparent and efficient flow of risk information among stakeholders, enhancing the rigor and timeliness of decision-making. However, the profound and systematic impacts of digital transformation on project management remain to be thoroughly explored in the existing literature. Based on this, we propose the following hypothesis of the research framework:

Hypothesis 1: 

Digital transformation in the construction industry has a significant and positive impact on the effectiveness of project risk management.

3.2. The Mediating Role of Environmental Governance Capacity

Environmental governance refers to the capacity to manage environmental impacts through policies, monitoring, and training [2]. Digital technologies can empower environmental simulation [24], enable real-time emissions monitoring [2], and facilitate data integration [24], which enhances governance foresight. Robust environmental governance reduces compliance risks, minimizes project delays, and enhances societal trust [2], thereby serving as a critical mediating pathway.

Environmental governance capacity represents an organization’s institutionalized and systematic capability to establish and implement environmental policies, objectives, management systems, training programs, and emergency response mechanisms. Digital transformation empowers and elevates this capacity through multiple channels. Technologically, digital infrastructure advancements, such as environmental monitoring sensor networks and energy management platforms, enable the real-time, precise collection, processing, and analysis of environmental performance data [2]. In terms of organization, the shift in digital strategy and cultural transformation encourages enterprises to adopt sustainability and environmental protection as core values and operational norms [25].

Enhanced environmental governance directly contributes to project risk management. Enterprises with robust environmental management systems can systematically and proactively identify environmental compliance, excessive resource consumption, and pollution incident risks throughout a project’s lifecycle [4,16]. This advantage allows them to establish standardized prevention and response protocols. Integrating digital environmental data streams with project risk management platforms enables environmental risks to be assessed and monitored alongside other engineering and financial risks, thereby enhancing the comprehensiveness and precision of risk management. Therefore, we propose:

Hypothesis 2: 

Environmental governance capacities mediate the relationship between digital transformation in the construction industry and project risk management levels.

3.3. The Mediating Role of Green Innovation Efficiency

Green innovation efficiency refers to the ability to develop solutions that deliver environmental and economic value simultaneously [2]. Digital technologies can accelerate green innovation by facilitating virtual prototyping [18], digital twin testing [2], and the use of knowledge-sharing platforms [5]. Highly efficient green innovation can enhance climate resilience, mitigate supply chain risks, and align with regulatory trends [26,27] to indirectly elevate risk management capabilities.

Green innovation efficiency is a company’s ability to reduce resource consumption, minimize environmental pollution, and maximize ecological benefits by developing or adopting new technologies, processes, products, or management models. Digital transformation is a key driver of green innovation. Digital technologies such as BIM optimize building form and envelope design to achieve energy savings at the source. At the same time, the Internet of Things (IoT) and artificial intelligence (AI) enable intelligent energy management and adaptive control in buildings. Furthermore, digital transformation enhances the efficiency of green innovation by facilitating knowledge flows, reducing trial-and-error costs, and encouraging companies to pursue green innovation [28].

Improving the efficiency of green innovation means that projects gain access to more advanced, environmentally friendly, and efficient technological solutions. This measure proactively manages technical, operational, and compliance risks associated with increasingly stringent carbon-emission regulations. For example, adopting digital, design-based, prefabricated, and modular construction methods substantially reduces on-site material waste and debris, while also enhancing construction quality and speed. These methods significantly mitigate traditional risks such as project delays, cost overruns, and safety incidents [29]. Therefore, we propose:

Hypothesis 3: 

Green innovation efficiency mediates the relationship between digital transformation in the construction industry and project risk management levels.

3.4. The Regulatory Role of Labor Investment Efficiency

Labor investment efficiency measures the effectiveness of human resource allocation relative to optimal levels [30]. Highly efficient organizations can accurately identify skill gaps, provide targeted training, and redeploy personnel rapidly to address risks [13,31].

Labor investment efficiency measures the extent to which enterprises optimize human resource allocation, specifically whether their hiring, retention, and deployment decisions deviate from optimal levels based on economic fundamentals [30]. High labor investment efficiency enables enterprises to precisely assign employees with suitable digital skills, risk management knowledge, and green innovation awareness to critical positions in digital transformation and project execution, ensuring the necessary technical and managerial talent is available for digital transformation, thereby enhancing its direct impact on risk management levels. Conversely, low labor efficiency weakens this effect. Therefore, we propose:

Hypothesis 4: 

Labor investment efficiency positively moderates the relationship between digital transformation and project risk management. In other words, the positive impact of digital transformation on project risk management levels is more pronounced when labor investment efficiency is higher.

4. Research Methodology

4.1. Sample Selection

This study employs a quantitative empirical approach, focusing on listed construction companies in China’s A-share market to investigate the impact mechanism of digital transformation on project risk management levels.

Sample selection adheres to the following principles: (1) the sample is restricted to construction enterprises continuously listed between 2012 and 2023 to ensure data continuity and comparability; (2) companies with financial anomalies (ST/ST*) and observations with over 30% missing data were excluded. The final panel dataset comprises 50 firms and 434 valid observations.

Data primarily sourced from the China Stock Market & Accounting Research (CSMAR) Database, Chinese Research Data Services (CNRDS) Database, and Corporate Social Responsibility (CSR) Reports, with key variables manually collected and validated to ensure accuracy and completeness.

4.2. Variable Measurement

Based on the research theme of this article, the following are the selection and measurement methods of each research variable (see

Table 1. Selection and definition of main research variables.

Variable		Description
Panel A: Baseline model
Dependent variable	PRML	Project risk management level
Independent variable	DTI	Digital transformation intensity
Panel B: Mechanism tests
EGC		Environmental governance capacity
GIE		Green innovation efficiency
LIE		Labor investment efficiency
Panel C: Control variables
DERI		Default risk
ROA		Return on assets
CAIN		Capital intensity = Fixed assets/total assets
SIZE		Enterprise size, equal to the natural logarithm of total assets.
DPI		Dual position integration means that one person holds both the chairperson and general manager positions.
EBD		Equity balance degree refers to the ratio of the shareholding proportions of the second to fifth-largest shareholders to that of the largest shareholder.
MSR		Managerial shareholding ratio
Firm		Firm dummy
Year		Year dummy

).

4.2.1. Dependent Variable: Project Risk Management Level (PRML)

To quantify the project risk management level (PRML), this study employs a systematic textual analysis of firms’ annual financial reports and Corporate Social Responsibility (CSR) disclosures. The measurement framework is grounded in the conceptual dimensions proposed by Kassem & Ahmed (2022) [20] and is contextualized to reflect the specific regulatory disclosure environment of Chinese construction enterprises. Specifically, PRML is assessed across three theoretically derived dimensions: (1) the comprehensiveness of the risk management system, (2) the timeliness of risk identification, and (3) the effectiveness of risk response measures.

The construction of the PRML index follows a four-step protocol:

• Dictionary Development: We establish a specialized dictionary of risk management terminology relevant to the construction sector.
• Data Collection: Python 3.14 web-crawling technology is utilized to extract the relevant textual content from the annual reports and CSR disclosures of the sample firms.
• Text Processing: The collected texts are processed using the Jieba Chinese word segmentation library. This step identifies and aggregates the frequencies of keywords associated with four critical risk domains: operational management risks, technological innovation risks, occupational safety risks, and environmental risks.
• Quantification: The index is calculated as the natural logarithm of one plus the total frequency of risk-related terms. This logarithmic transformation mitigates the right-skewness commonly found in textual frequency distributions [32] while preserving the relative intensity of risk disclosure [33].

The approach provides an objective, replicable metric that captures both the depth and strategic emphasis of construction corporate risk management practices within China’s institutional context.

To address concerns regarding common-source bias and to validate that the textual metric reflects actual risk management practices rather than “impression management”, we conducted an external criterion validation [34,35]. We matched the sample with the Chinese Enterprise Environmental Administrative Penalty Database issued by CnOpenData. The results show a statistically significant negative correlation between PRML and the frequency of environmental penalties ($\lambda$ = −0.096, p < 0.05). This discriminant validity evidence confirms that firms disclosing higher PRML scores are objectively less likely to violate regulations, indicating that the textual measure captures substantive risk control capabilities rather than rhetorical embellishment.

4.2.2. Independent Variable: Digital Transformation Intensity (DTI)

Corporate digital capability refers to the process by which enterprises convert digital technology resources into dynamic capabilities; the strength of this capability determines the potential of enterprises to respond to environmental uncertainties. Therefore, this paper employs digital transformation intensity (DTI) obtained through text analysis to capture the degree to which digital resources are internalized into strategic decision-making processes, thereby reflecting the enterprise’s ability to absorb digital technologies.

This study constructs a firm-level digital transformation intensity (DTI) by systematically analyzing the “Management Discussion and Analysis” (MD&A) sections of corporate annual reports. The measurement protocol follows the text analysis framework established by Zhai et al. (2022) [24], which is subsequently adapted to reflect the technology-centric trajectory specific to the construction industry.

The specific construction steps are as follows:

• Dictionary Development: We first establish a specialized dictionary of digital terminology relevant to construction enterprises.
• Data Collection: Python web-crawling technology is employed to collect the MD&A content from the sample firms’ annual financial reports.
• Text Processing: Utilizing the Jieba Chinese word segmentation library, the collected texts undergo segmentation and filtering. This process identifies occurrences of predefined keywords related to digital technology applications, including Big Data, Cloud Computing, the Internet of Things (IoT), Artificial Intelligence (AI), Building Information Modeling (BIM), and Digital Twin technologies.
• Quantification: The index is generated through frequency statistics of these keywords, followed by a logarithmic transformation to mitigate skewness.

The dictionary-based text analysis approach offers a contextualized and relatively objective metric for assessing corporate digital technology engagement. By capturing the strategic discourse of firms concerning technology integration, this method can partially address the limitations of narrow scope and endogeneity commonly associated with traditional financial proxy variables, thereby contributing to enhanced variable validity within the construction industry context.

To establish the construct validity of DTI and rule out the possibility that it merely captures general “disclosure verbosity”, we utilized objective financial data as an external criterion. Following Anderson et al. (2025) [36], we extracted the proportion of intangible assets related to digital technology as “digital intangible assets” from the “Notes to Financial Statement”. The correlation analysis reveals a significant positive relationship between textual DTI and objective digital intangible assets investment ($\lambda$ = 0.128, p < 0.05). This convergent validity evidence demonstrates that the textual frequency of digital keywords aligns with capital allocation, confirming that DTI measures real strategic resource internalization.

Furthermore, the distinct construct connotations and measurement dimensions of PRML and DTI help to mitigate concerns regarding artificial correlations arising from measurement overlap. Specifically, PRML measures the sophistication of the risk management system (including identification, response, and control mechanisms), emphasizing the “soft power” of institutional development. In contrast, the core logic of DTI lies in capturing the breadth of enterprises’ application of digital technology tools (such as Big Data, AI, and Cloud Computing), focusing on the “hard power” of technology adoption. The two represent different aspects of “compliance management” and “technological innovation” in corporate strategy, respectively, with distinct underlying lexicons and theoretical logics. While this conceptual and measurement independence reduces the likelihood of spurious correlations caused by semantic similarity, it does not fully rule out endogeneity concerns such as reverse causality and omitted variable bias. We employ a difference-in-differences (DID) model in Section 5.5.4, which provides more robust causal identification.

4.2.3. Mechanism Variables

• Environmental governance capacity (EGC). EGC was operationalized leveraging China’s mandatory environmental disclosure regime, which requires listed firms to report environmental governance practices in annual financial reports, CSR statements, and sustainability disclosures. Following the methodological framework of Li et al. (2023) [2], we extracted eight standardized environmental disclosure items (X₁–X₈; operational definitions in

  Table 2. Items constituting the environmental governance capacity (EGC).

  Item	Description
  X1	Environmental protection concepts
  X2	Environmental protection goals
  X3	The environmental protection management system
  X4	Environmental protection education and training
  X5	Special environmental protection campaigns
  X6	Emergency response mechanisms for environmental incidents
  X7	Environmental protection honors or awards
  X8	“Three Simultaneities” system

  ) from these documents via the CSMAR database. A composite EGC score was generated through factor analysis. This metric holistically captures firms’ environmental stewardship from regulatory compliance and pollution control to proactive sustainability initiatives, offering a multidimensional, disclosure-based assessment that aligns with institutional expectations in China’s evolving green governance landscape.
• Green innovation efficiency (GIE). Owing to the absence of a universally standardized metric for green innovation, measurement approaches remain context-dependent across scholarly domains. To ensure construct validity and contextual relevance, this study defines GIE by manually identifying and verifying green invention patent grants using the official green patent classification scheme issued by the China National Intellectual Property Administration (CNIPA). Following established methodological precedents [37,38], we prioritized green invention patents over utility model patents, as the former undergo rigorous substantive examination and more credibly reflect firms’ capacity for high-quality, technologically substantive environmental innovation. The GIE index was computed as ln(green invention patent grants + 1) to capture the depth and quality of firms’ environmentally oriented innovation output within China’s institutional framework.
• Labor investment efficiency (LIE). LIE was measured as the absolute value of residuals from an expected labor investment model, adapted from established methodologies [30,39,40]. First, we construct the expected labor investment model:

$${HIRE}_{it}={\rho }_0+{\rho }_1{SALEGROTH}_{i,t-1}+{\rho }_2{SALEGROTH}_{it}+{\rho }_3{∆ROA}_{it}+{\rho }_4{∆ROA}_{i,t-1}$$

$$+{\rho }_5{ROA}_{it}+{\rho }_6{RETURN}_{it}+{\rho }_7{VALUE}_{it}+{\rho }_8{QUICK}_{i,t-1}+{\rho }_9{∆QUICK}_{i,t-1}$$

$$+{\rho }_{10}∆{QUICK}_{it}+{\rho }_{11}{LEV}_{i,t-1}+{\rho }_{12}{LOSSBIN1}_{i,t-1}+{\rho }_{13}{LOSSBIN2}_{i,t-1}$$

$$+{\rho }_{14}{LOSSBIN3}_{i,t-1}+{\rho }_{15}{LOSSBIN4}_{i,t-1}+{\rho }_{16}{LOSSBIN5}_{i,t-1}+{\epsilon }_{it}$$

(1)

among these, HIRE represents the rate of change in the number of enterprise employees, referred to as labor investment; ∆ denotes the change value; SALEGROTH is the sales growth rate; ROA is the return on assets; RETURN is the annual stock return rate; VALUE is the natural logarithm of the enterprise market value; QUICK is the quick ratio; and LEV is the debt-to-asset ratio; Considering that Pinnuck & Lillis (2007) [41] argue that loss-making firms are more likely to lay off employees compared to profitable firms, this paper divides the range of ROA from −0.156 to 0 into five intervals. LOSSBIN1–LOSSBIN5 are defined based on the loss ranges of ROA. For instance, if ROA falls within the first interval [−0.0312, 0], then LOSSBIN1 = 1; otherwise, it is 0.

Subsequently, labor investment efficiency (LIE) is defined as the negative of the absolute value of the residuals (ε) obtained from the regression of Equation (1). A higher value indicates greater LIE, and conversely, a lower value signifies lower LIE.

4.2.4. Control Variables

Following the approach of Mändli & Rönkkö (2025) [42], this study controls for enterprise characteristic variables that may affect project risk management level: (1) default risk (DERI), using the Naïve model proposed by Bharath & Shumway (2008) [43] to estimate default probability as a proxy variable for DERI; (2) profitability (ROA), measured by return on total assets; and (3) capital intensity (CAIN), measured by the ratio of fixed assets to total assets; (4) firm size (SIZE), defined as the natural logarithm of total assets. Additionally, corporate governance variables are selected: (5) dual positions integration (DPI), a binary indicator (1 = chairperson and CEO positions are the same individual); (6) equity balance degree (EBD), defined as the ratio of the combined shareholding of the second to fifth largest shareholders to that of the largest shareholder; and (7) managerial shareholding ratio (MSR), reflecting the proportion of shares held by senior executives. Finally, Firm fixed effects are selected to control the impact of individual trends; Year fixed effects are included to absorb macroeconomic fluctuations and temporal trends. All variable definitions are summarized in Table 1.

4.3. Empirical Models

To systematically test the research hypotheses, four panel regression models were specified. The baseline model (Equation (2)) first assesses the direct effect of DTI on PRML, corresponding to Hypothesis 1:

$${PRML}_{it}={\alpha }_0+{\alpha }_1{DTI}_{it}+{\theta }^{\prime }\sum {Controls}_{it}+\sum Firm+\sum Year+{\epsilon }_{it}.$$

(2)

To deeply explore the inherent logical chain and “black box” operational mechanism of how digital technology application affects project risk management, based on the approaches proposed by Baron & Kenny (1986) [44] and Wang et al. (2024, 2026) [45,46], this study introduces environmental governance capacity (EGC) and green innovation efficiency (GIE) as mediating variables. On the basis of the existing Equation (1), we constructed Equations (3)–(6) for mediating effect analysis to respectively test the mediating effect of DTI on PRML via EGC (this part corresponds to Hypothesis 2), and the effect of DTI on PRML via GIE (this part corresponds to Hypothesis 3).

First, Equations (3) and (4) are employed to verify whether there is a significant association between the independent variable and each mediating variable. Second, Equations (5) and (6) are used to examine the impact of the independent variable and mediating variables on the dependent variable. Finally, the mediating effects of the independent variable on the dependent variable through the mediating variables are calculated.

$${EGC}_{it}={\beta }_0+{\beta }_1{DTI}_{it}+{\theta }^{\prime }\sum {Controls}_{it}+\sum Firm+\sum Year+{\epsilon }_{it},$$

(3)

$${GIE}_{it}={\beta }_0+{\beta }_1{DTI}_{it}+{\theta }^{\prime }\sum {Controls}_{it}+\sum Firm+\sum Year+{\epsilon }_{it},$$

(4)

$${PRML}_{it}={\beta }_0+{\beta }_1{DTI}_{it}+{\beta }_2{EGC}_{it}+{\theta }^{\prime }\sum {Controls}_{it}+\sum Firm+\sum Year+{\epsilon }_{it};$$

(5)

$${PRML}_{it}={\beta }_0+{\beta }_1{DTI}_{it}+{\beta }_2{GIE}_{it}+{\theta }^{\prime }\sum {Controls}_{it}+\sum Firm+\sum Year+{\epsilon }_{it}.$$

(6)

To rigorously examine the boundary condition role of labor investment efficiency (LIE) in the relationship between DTI and PRML, and to strengthen the theoretical foundation of causal inference, this study positions the moderation analysis at the core of causal mechanism validation. Grounded in the RBV [10] and dynamic capabilities theory [11], LIE is an objective measure of human resource allocation efficiency and serves as a key organizational capability prerequisite for translating digital technology into essential risk governance outcomes. This moderating pathway not only verifies the theoretical hypothesis but also enhances internal validity by revealing the conditional dependence of the causal relationship, thereby mitigating concerns about spurious correlations driven by industry-wide trends or macro-environmental shocks [47]. The moderated regression model is specified as follows:

$${PRML}_{it}={\gamma }_0+{\gamma }_1{DTI}_{it}+{\gamma }_2{LIE}_{it}+{\gamma }_3{DTI}_{it}\times {LIE}_{it}+{\theta }^{\prime }\sum {Controls}_{it}$$

$$+\sum Firm+\sum Year+{\epsilon }_{it},$$

(7)

where the denotation of each variable in Equations (2)–(7) is shown in Table 1. $\epsilon$ is a random error term. Controls are the collective term for control variables.

5. Results

5.1. Factor Analysis Results of Environmental Governance Capacity (EGC)

To evaluate the measurement quality of the EGC scale for system validation, this study sequentially conducts the following tests:

• Reliability analysis: The scale exhibited strong internal consistency, with a Cronbach’s α coefficient of 0.773 (see

  Table 3. The reliability test results of factor analysis for EGC.

  Item	Obs	Sign	Item-Test Correlation	Item-Rest Correlation	Average Interitem Correlation	Cronbach’s $\alpha$
  X1	434	+	0.645	0.502	0.293	0.743
  X2	434	+	0.720	0.599	0.275	0.726
  X3	434	+	0.730	0.612	0.272	0.724
  X4	434	+	0.708	0.583	0.278	0.729
  X5	434	+	0.683	0.550	0.284	0.735
  X6	434	+	0.516	0.345	0.323	0.770
  X7	434	+	0.560	0.397	0.313	0.761
  X8	434	+	0.408	0.220	0.349	0.789
  Test scale					0.298	0.773

  ), meeting established psychometric thresholds [48]. Removal of any single item did not elevate the α coefficient, confirming that all items stably and positively contribute to the scale’s structural reliability.
• Validity assessment: The KMO measure reached 0.831, and Bartlett’s test of sphericity was highly significant (${\chi }^2$ = 795.44, df = 28, p-value < 0.001; see

  Table 4. The validity test results of factor analysis for EGC.

  Testing Indicators	Statistics
  Bartlett test of sphericity	${\chi }^2$ = 795.44, df = 28 p-value = 0.000
  KMO	0.831

  ), supporting data suitability for factor analysis.
• Scree plot examination: The scree plot (Figure 2) displayed a clear inflection point after the second eigenvalue, corroborating the extraction of two factors in alignment with both empirical criteria and theoretical expectations.
• Variance explained after rotation: Following varimax rotation, the two-factor solution cumulatively accounted for 53.87% of total variance (see

  Table 5. Rotated variance contribution of each factor.

  Principle Factor	Eigenvalue	Percentage of Variance Contribution	Percentage of Cumulative Variance Contribution
  F1	2.818	0.3523	0.3523
  F2	1.491	0.1864	0.5387

  Note: Method: principal-component factors.

  ). By further combining Table 2 and

  Table 6. Results of each item score after rotation.

  Item	F1	F2
  X1	0.246	−0.037
  X2	0.156	0.203
  X3	0.163	0.197
  X4	0.250	0.007
  X5	0.324	−0.163
  X6	−0.070	0.487
  X7	0.323	−0.270
  X8	−0.189	0.619

  , it can be observed that Factor 1 (F₁) explained 35.23% of variance, reflecting institutional dimensions (X₁–X₅, and X₇); and Factor 2 (F₂) explained 18.64% of variance, capturing operational responsiveness (X₆ and X₈).

5.2. Descriptive Statistics

Table 7. Descriptive statistics of main research variables.

Variable	N	Median	Mean	SD	Min	Max
Panel A: Primary variables
PRML	434	0.693	0.683	0.343	0	1.386
DTI	434	0.693	0.733	0.920	0	5.799
Panel B: Mechanism variables
EGC	434	−0.079	0	0.399	−0.412	0.920
GIE	434	1.099	1.705	1.758	0	6.615
LIE	434	−0.174	−0.231	0.278	−2.84	0
Panel C: Control variables
DERI	434	0	0.102	0.302	0	1
ROA	434	0.022	0.026	0.033	−0.156	0.121
CAIN	434	1.849	2.738	2.935	0.760	22.715
SIZE	434	23.852	24.225	2.275	19.585	28.697
DPI	434	0	0.150	0.358	0	1
EBD	434	0.459	0.590	0.558	0.008	3.208
MSR	434	0.011	5.045	11.802	0	53.598

The italics are variables and defined as indicated in Table 1.

summarizes the distributional characteristics of key study variables. PRML averaged 0.683 (SD = 0.343; range: 0–1.386), indicating moderate heterogeneity in risk management maturity across construction firms. DTI showed a mean of 0.733 (SD = 0.920) with a right-skewed distribution (maximum = 5.799), reflecting pronounced divergence in digital technology adoption.

Among mechanism variables, EGC, a composite score derived from factor analysis, centered at 0 (SD = 0.399); GIE averaged 1.705 (SD = 1.758), revealing substantial cross-firm variation in green innovation output; LIE averaged 0.231 (SD = 0.278), with right-skewness suggesting human resource over-investment in certain firms, consistent with Khedmati et al. (2020) [30].

5.3. Model Selection Tests

Guided by the statistical diagnostics in

Table 8. Selective testing of the baseline model.

Statistic	Multivariate Mixed Model	Random Effects Panel Model	Fixed Effects Panel Model
B-P test (p-value)	${\overline{\chi }}^2$ = 433.27 (0.000)
Hausman test (p-value)		${\chi }^2$ = 118.49 (0.000)

, this study implemented a rigorous model selection protocol to identify the optimal estimation strategy. The B-P test strongly rejected the null hypothesis of no individual effects (${\overline{\chi }}^2$ = 433.27), confirming significant unobserved heterogeneity across firms and thereby ruling out the multivariate mixed model.

The subsequent Hausman specification test yielded a highly significant statistic (${\chi }^2$ = 118.49), indicating systematic correlation between individual-specific effects and core explanatory variables. This result confirms that the random effects estimator would produce inconsistent estimates, whereas the fixed effects estimator remains consistent.

Accordingly, a two-way fixed effects panel model, simultaneously controlling for firm-specific and year fixed effects, was adopted. This specification effectively isolates time-invariant unobserved heterogeneity and macro-temporal shocks, which is particularly appropriate given the pronounced firm-specific dependence embedded in construction enterprises’ risk profiles.

5.4. Baseline Model and Mechanism Test Results

Table 9. Test results on the benchmark models of DTI on PRML and their impact mechanism analyses.

Variable		PRML	Mediating Effects				Moderating Effect
			EGC	PRML	GIE	PRML	PRML
DTI		0.045 ** (0.023)	0.044 ** (0.021)	0.043 ** (0.018)	0.178 ** (0.076)	0.050 ** (0.023)	0.055 ** (0.019)
EGC				0.136 ** (0.056)
GIE						0.038 ** (0.015)
LIE							0.114 ** (0.053)
DTI × LIE							0.143 ** (0.065)
DERI		−0.140 *** (0.054)	−0.016 (0.051)	−0.128 ** (0.056)	−0.398 ** (0.188)	−0.117 ** (0.056)	−0.127 ** (0.057)
ROA		−0.408 (0.479)	0.306 (0.416)	−0.134 (0.465)	0.197 (1.523)	−0.120 (0.465)	−0.308 (0.485)
CAIN		−0.001 *** (0.009)		0.003 (0.009)		0.004 (0.009)	0.003 (0.009)
SIZE			0.139 *** (0.018)	−0.014 (0.021)	0.811 *** (0.064)	−0.021 (0.023)	−0.042 (0.028)
DPI		0.010 (0.044)		0.031 (0.044)		0.012 (0.045)	−0.006 (0.044)
EBD			−0.012 (0.040)	−0.064 (0.043)	−0.105 (0.145)	−0.065 (0.043)	−0.098 ** (0.044)
MSR			−0.004 (0.003)	0.002 (0.004)	−0.032 ** (0.013)	0.002 (0.004)	0.005 (0.039)
Firm		Yes	Yes	Yes	Yes	Yes	Yes
Year		Yes	Yes	Yes	Yes	Yes	Yes
cons		0.700 *** (0.067)	−3.390 *** (0.426)	1.021 ** (0.499)	−17.537 *** (1.562)	1.113 ** (0.535)	1.662 ** (0.647)
N		434	434	434	434	434	434
${\mathrm{R}}^2$		8.52%	17.14%	5.38%	33.78%	4.82%	11.82%
F-value		2.14	13.03	2.37	32.14	2.38	2.32
Bootstrap Sobel test	Direct-effect		0.0429 ** (0.018) [0.0083, 0.0775]		0.0496 ** (0.019) [0.0277, 0.0469]
	Indirect-effect		0.0060 ** (0.004) [0.0001, 0.0003]		0.0068 ** (0.005) [0.0021, 0.0157]
Total effect			0.0489 ** (0.018)		0.0564 ** (0.019)

Note: **, *** indicate significance at the 0.05, 0.01 level, respectively. Standard errors are presented in parentheses. The italics are variables and defined as indicated in Table 1.

reports the estimation results of the baseline model and mechanism tests.

5.4.1. Baseline Model Test of DTI on PRML

The baseline regression (Column 1) indicates that DTI exerts a significant positive effect on PRML (${\alpha }_1$ = 0.045), supporting Hypothesis 1. This finding validates the core mechanism through which digital technologies enhance risk management efficacy via information integration and real-time monitoring [5]. Among control variables, DERI significantly and negatively affects PRML, aligning with theoretical expectations that financial constraints undermine risk response capacity.

In addition, the control variables of the baseline model have played effective roles in stabilizing the model and improving the universality of the research results. In particular, the significant effect of DERI at the 0.01 level is −0.140, indicating that project risk management should strengthen the control and management of default risk.

5.4.2. Mediating Effect Tests

• Environmental governance capacity (EGC) pathway: The results in Column 2 demonstrate a significant positive impact of DTI on EGC (${\beta }_1$ = 0.044), indicating that digital transformation effectively strengthens corporate capabilities in environmental policy implementation, monitoring systems, and emergency response. This finding aligns with Li et al. (2023) [2], who corroborated that digital infrastructure and sensor networks facilitate real-time, refined environmental performance management. Crucially, Column 3 reveals that EGC also exerts a significant positive influence on PRML (${\beta }_2$ = 0.136). This result suggests that robust environmental governance capabilities enable the systematic identification of compliance, resource consumption, and pollution risks throughout the project lifecycle, thereby enhancing overall risk management, consistent with Chen et al. (2024) [4] and Oke et al. (2023) [15]. Furthermore, to ensure robustness, we employed the bootstrap method (1000 resample), and the results indicate that the indirect effect of DTI on PRML through EGC is 0.006, with a 95% bias-corrected confidence interval (CI) of [0.0001, 0.0003]. Since the interval excludes zero, this provides evidence supporting the mediating role of EGC (Hypothesis 2).
• Green innovation efficiency (GIE) pathway: Column 4 indicates that DTI has a significant positive effect on GIE (${\beta }_1$ = 0.178). This result supports the findings of Tian et al. (2022) [18], suggesting that technologies such as BIM, IoT, and AI accelerate green technology development through virtual prototyping, digital twin testing, and knowledge-sharing platforms. Concurrently, the analysis results in Column 5 show that GIE significantly positively impacts PRML (${\beta }_2$ = 0.030), which confirms that higher green innovation efficiency optimizes project risk management, as noted by Li et al. (2023) [2]. Digitalization-driven prefabrication and modular construction reduce material waste and carbon compliance risks while minimizing safety incidents and delays through process optimization. Similarly, the bootstrap results show a significant indirect effect through GIE (Effect = 0.0068, 95% CI [0.0021, 0.0157]). These findings suggest that GIE serves as a viable mechanism linking DTI to PRML, thus supporting Hypothesis 3.

5.4.3. Moderating Effect Test

• Boundary condition of labor investment efficiency (LIE): Column 6 presents the moderating analysis. The interaction term DTI × LIE is significantly positive (${\gamma }_3$ = 0.143), confirming that the positive effect of DTI on PRML is strengthened under higher labor investment efficiency (i.e., higher LIE values). Simple slope analysis (Figure 3) clarifies this boundary condition: under low-efficiency conditions (${LIE}^L$ = −0.509 = ${Mean}_{LIE}-1{SD}_{LIE}$), DTI exhibits a negative marginal effect on PRML (${\gamma }_1+{\gamma }_3{LIE}^L$ = −0.018); under high-efficiency conditions (${LIE}^H$ = 0.047 = ${Mean}_{LIE}+1{SD}_{LIE}$), the effect turns positive (${\gamma }_1+{\gamma }_3{LIE}^H$ = 0.062). A critical threshold is identified at LIE = −0.385 ($-{\gamma }_1/{\gamma }_3$): when LIE falls below this value, digital transformation fails to enhance risk management efficacy (marginal effect becomes negative); conversely, only when LIE exceeds this threshold does DTI significantly and positively influence PRML. This finding not only lends support to Hypothesis 4 but also underscores a pivotal practical insight: precise human resource allocation, not merely technological adoption depth, is a necessary precondition for unlocking the risk governance value of digital technologies. Human resource misallocation substantially diminishes technological efficacy, corroborating the sociotechnical systems theory tenet of “technology–organizational capability” co-evolution [7]. The quantifiable threshold further offers actionable guidance for construction enterprises implementing digital transformation initiatives.

5.5. Robustness Tests

To ensure the reliability of core findings from the baseline model (Equation (2)), this study implements a tripartite robustness testing strategy: common-source bias test of PRML and DTI, key variable replacement, heterogeneity analysis by nature of property rights, and a quasi-natural experimental design (results summarized in

Table 10. Results of the robustness test of the baseline model.

Variable	(1) Common-Source Bias Test	(2) Replace Variables	(3) Heterogeneity Test		(4) DID Test
	PRML	PRML	State-Controlled	Non-State-Controlled	PRML
DTI	0.045 ** (0.023)		0.046 (0.034)	0.077 ** (0.038)
ADTI		0.045 ** (0.023)
Time					0.034 (0.051)
Treat					−0.171 (0.112)
Time × Treat					0.255 ** (0.117)
MD&A	−0.043 (0.047)
DERI	−0.158 *** (0.054)	−0.140 *** (0.054)	−0.080 (0.113)	−0.164 ** (0.073)	−0.146 *** (0.055)
ROA	−0.480 (0.474)	−0.408 (0.479)	0.603 (0.876)	−1.261 (0.666)	0.412 (0.511)
CAIN	−0.001 (0.009)	−0.001 *** (0.009)	0.008 (0.010)	−0.041 (0.022)	−0.093 ** (0.027)
SIZE					−0.010 (0.008)
DPI	0.012 (0.044)	0.010 (0.044)	−0.007 (0.076)	0.023 (0.063)	0.008 ** (0.003)
Firm	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes
cons	1.109 ** (0.442)	0.637 *** (0.040)	0.684 *** (0.072)	0.756 *** (0.278)
N	434	434	434	434	434
${\mathrm{R}}^2$	9.29%	8.52%	8.67%	15.48%	8.51%
F-value	2.18	2.14	3.32	3.47	4.94

Note: **, *** indicate significance at the 0.05, 0.01 level, respectively. Standard errors are presented in parentheses. The italics are variables and defined as indicated in Table 1.

).

5.5.1. Common-Source Bias Test

To further isolate the impact of potential common-source bias between PRML and DTI arising from “disclosure style”, we control for the total count of the MD&A section in annual financial reports (MD&A, log-transformed), drawing on the ideas of Podsakoff et al. (2003) [34] and Faiz et al. (2024) [49]. If the regression results of the baseline model (Equation (2)) are driven solely by firms that produce longer reports, the impact of DTI on PRML would no longer be significant after controlling for length. Column 1 of Table 10 reports the test results, showing that the impact of DTI on PRML remains significantly positive (${\alpha }_1$ = 0.045) after controlling for MD&A, indicating no common-resource bias in the PRML and DTI constructed in this study.

5.5.2. Key Variable Replacement Test

To address concerns regarding measurement specificity, this study re-examines the relationship using alternative digital transformation intensity (ADTI). ADTI strictly follows the construction framework of Zhai et al. (2022) [24], derived from full-text mining of listed companies’ annual reports. It systematically identifies keyword frequencies across four technological dimensions (such as artificial intelligence, blockchain, cloud computing, and big data) and technology application scenarios. The composite index is generated through term frequency normalization and industry-year two-dimensional winsorization. Regression results in Column 2 of Table 10 indicate that substituting DTI with ADTI yields a consistently significant positive effect on PRML (${\alpha }_1$ = 0.038), with the coefficient direction and statistical significance closely aligning with the baseline model. This finding not only reinforces the robustness of Hypothesis 1 but also substantially enhances theoretical validity through a multi-source construct measurement strategy, effectively mitigating conclusion bias attributable to indicator specificity.

5.5.3. Heterogeneity Analysis by Nature of Property Rights

Considering the moderating role of institutional environments on technology adoption, subgroup analyses of the baseline model (Equation (2)) are conducted based on ultimate controlling ownership (state-controlled vs. non-state-controlled enterprises). Results in Columns 3 and 4 of Table 10 reveal that the positive effect of digital transformation on PRML is significantly stronger in non-state-controlled enterprises (${\alpha }_1$ = 0.077). In contrast, the effect in state-controlled enterprises is not statistically significant (${\alpha }_1$ = 0.046). This divergence stems from institutional logic heterogeneity: non-state-controlled enterprises, facing heightened market competition, demonstrate more agile responsiveness to technology-enabled empowerment; conversely, state-controlled enterprises encounter greater resistance to management restructuring during digital transformation due to administrative constraints and hierarchical decision-making processes [50].

5.5.4. Quasi-Natural Experiment: Difference-in-Differences (DID) Analysis

To strengthen causal inference, this study implements a DID approach within a quasi-experimental framework. The selection of 2015 as the policy shock year is substantiated by three historically grounded rationales: (1) the State Council’s Action Plan for Promoting Big Data Development (August 2015) explicitly designated the construction industry as a priority sector for digital transformation; (2) the Ministry of Housing and Urban-Rural Development’s Guiding Opinions on Promoting the Application of Building Information Modeling (2015) mandated full BIM adoption in state-funded projects by 2017, with anticipatory implementation effects already observable in 2015; and (3) international evidence on the digital inflection point in construction [51] confirms that global BIM penetration surpassed a critical threshold (32%) in 2015, catalyzing technology adoption network effects.

Accordingly, we define the treatment group (listed construction firms that had initiated digital projects before 2015; Treat = 1) and the control group (firms without such initiatives; Treat = 0), and estimate the following model:

$${PRML}_{it}={\alpha }_0+{\beta }_1{Treat}_i+{\beta }_2{Post}_t+\gamma \left({Treat}_i\times {Post}_t\right)+{\theta }^{\prime }\sum {Controls}_{it}$$

$$+\sum Firm+\sum Year+{\epsilon }_{it},$$

(8)

where Post is a time dummy equal to 1 for Year ≥ 2015 (reflecting the policy implementation timeline for construction industry digitalization) and 0 otherwise; ${Treat}_i\times {Post}_t$ is the core interaction term, with coefficient $\gamma$ representing the DID estimator of the net policy effect. $Controls$ are covariates, defined consistently with the aforementioned models. DID estimation in Column 5 of Table 10 reveals that the treatment group exhibited a statistically significant increase of 25.5% in PRML following policy implementation, corroborating a causal relationship between digital practice adoption and risk management enhancement. This finding provides robust external validation for Hypothesis 1 from a policy intervention perspective and effectively mitigates endogeneity concerns.

Figure 4 presents the event study results, which support the parallel trend assumption. Using the year before policy implementation (2014) as the reference category, the pre-treatment coefficients (2012 and 2013) are statistically insignificant. Specifically, the coefficient for 2012 is close to zero, with its 95% confidence interval crossing the zero line. Although the 2013 coefficient is slightly positive, its confidence interval still includes zero, indicating no statistical significance. These results suggest that before the digital transformation initiative, there were no significant differences in risk management levels between the treatment and control groups, thereby satisfying the prerequisite for the DID model.

Collectively, the three robustness strategies demonstrate remarkable consistency: irrespective of variations in measurement approaches, institutional attributes, or estimation frameworks, the core relationship, “digital transformation → risk management enhancement”, remains empirically robust. Notably, the DID analysis leverages an exogenous policy shock to demonstrate that accelerated digitalization directly elevates risk management capabilities, yielding the strongest evidence for causal inference. These results not only reinforce theoretical validity but also illuminate how institutional contexts shape the boundary conditions of technology-enabled empowerment, thereby establishing a methodologically extensible foundation for future research.

6. Discussion

This study empirically reveals the dual-path mechanism and boundary conditions through which digital transformation in the construction industry improves project risk management. Its findings are multi-level, corroborated classical theories and global practices, deepening the understanding of the interactive logic of “technology–organization–environment”.

Sociotechnical systems theory emphasizes the co-evolution of technical subsystems and social subsystems [52]. This study finds that digital transformation empowers risk management through dual pathways: environmental governance capacity (with a mediating effect of 0.0060) and green innovation efficiency (with a mediating effect of 0.0068), which constitutes an empirical verification of this theory. Although the magnitudes of the mediating effects of the two paths are similar, their internal mechanisms of action differ significantly, which stems from the logical distinction between the institutional constraint path and the technology-driven path under the unique attributes of the construction industry.

From the perspective of mechanism attributes, the path of environmental governance capacity is mainly constrained by the rigidity of institutional implementation. At present, environmental regulations in the construction industry are still dominated by end-of-pipe treatment, and there is a contradiction between standardized regulations and project heterogeneity, such as the issue of adapting building energy efficiency standards to different climate zones. In addition, there may be a tendency towards local protectionism in environmental law enforcement, which leads to diminishing marginal returns on corporate investments in environmental governance, thereby limiting the scale of the mediating effect of this path.

In contrast, the path of green innovation efficiency involves interdisciplinary technological integration, such as R&D of low-carbon building materials, optimization of prefabricated construction techniques, and the coupled design of BIM and green buildings. Although the technological iteration cycle is long and risky, once a breakthrough is made, the resulting productivity improvement will have a multiplier effect. Furthermore, the effects of environmental governance (such as soil remediation) often exhibit significant assessment lags. In contrast, the achievements of green innovation (such as green building material certification) can be quickly monetized through market mechanisms to achieve premiums. This difference in the path driven by digital transformation through technology (GIE) (17.8%) is higher than that driven by institutions (EGC) (4.4%), yet their final mediating effects tend to be similar.

Therefore, the green transition of the construction industry should strive to build a two-way strengthening mechanism of technological breakthroughs and institutional adaptation: on the one hand, reduce institutional friction by dynamically adjusting environmental regulatory tools (such as introducing a carbon trading market mechanisms), and on the other hand, improve the green technology innovation ecosystem (such as establishing collaborative innovation platforms that deeply integrate industry, academia, research, and application) to overcome technical barriers, ultimately achieving the coordinated optimization of environmental and economic benefits. Technical tools such as BIM collision detection and IoT environmental sensing (the technical subsystems) must be embedded into environmental compliance processes and green R&D systems (the social subsystems) before they can be transformed into practical and effective risk management and control outcomes. This mechanism is particularly evident in China’s “Dual Carbon” policy practice: China Construction Eighth Engineering Division Corp., Ltd. (Shanghai, China) has applied the BIM-integrated carbon management platform to the Xiong’an Citizen Service Center project, simultaneously optimizing construction energy consumption and material recycling plans. This measure has not only strengthened environmental governance compliance but also mitigated resource shortages and policy compliance risks through early layout of green technological innovation. Internationally, UNStudio, a Dutch architectural firm, has deeply integrated digital twin technology with circular economy design, realizing real-time simulation of building material recycling pathways and carbon footprints, significantly reducing environmental risks across the entire life cycle, verifying the universality of green innovation efficiency as a core mediating pathway, and also echoes the viewpoint proposed by Chen et al. (2024) [4] that technological value needs to be transformed through organizational innovation.

The resource-based view holds that sustainable competitive advantage stems from the ability to integrate heterogeneous resources [10]. This study confirms that labor investment efficiency serves as a key moderating variable for unlocking the effectiveness of digital transformation: when human resource allocation is precise and efficient, the promotional effect of technology on risk management increases by 25.5%; the effect diminishes significantly in the case of misallocation. This result reveals that “digital technology” itself is not a scarce resource; rather, the organizational dynamic capability to convert technology into risk response capabilities [11] constitutes the core of value. Practices in China provide vivid illustrations: China Vanke has implemented a “digital job profile” mechanism, precisely assigning BIM engineers to design clash detection and construction early warning nodes, which has significantly improved risk response efficiency for the Shenzhen Zhenshan Mansion project. In contrast, some state-controlled enterprise projects with rigid human resource management have left advanced digital platforms idle due to skill mismatches despite their deployment. International cases further corroborate this logic: Surbana Jurong of Singapore established cross-functional digital teams for the Changi Airport Terminal 5 project, leveraging efficient human collaboration to transform digital twin technology into real-time risk intervention capabilities, which aligns closely with the “technology-organization fit” proposition emphasized by Chen et al. (2019) [7].

In summary, the findings of this study are not isolated discoveries, but engage in an organic verification with sociotechnical systems theory, the resource-based view, and institutional theory. Technological empowerment must be translated through organizational mechanisms (environmental governance and green innovation) and can only unleash its risk management value under suitable human resource conditions. China’s transitional practice serves as a reference for cases under diverse global institutional contexts, jointly verifying that the digital transformation of the construction industry is essentially a systematic collaboration of technological, organizational, and environmental factors, rather than mere technological superposition.

7. Conclusions and Prospects

Based on the sociotechnical system theory and the resource-based view, this study empirically reveals the core mechanism through which digital transformation in the construction industry empowers project risk management. Digital transformation is not an isolated technological application; instead, it delivers value via the dual paths of “environmental governance optimization” and “green innovation improvement”, and achieves maximum effectiveness under the boundary condition of “precise human resource allocation”. This research framework of “technology empowerment → governance optimization/innovation improvement → human collaboration” systematically addresses the core academic concern of “why technology investment has generally failed to translate into practical risk control effects”, providing a theoretical anchor and empirical evidence for the digital transformation of the construction industry to shift from “tool superposition” to “systematic collaboration”.

7.1. Management Insights: Implications for China and Reference for Emerging Economics

While the empirical evidence is drawn from Chinese listed firms, the insights derived offer a reference for construction industries in other emerging economies undergoing similar digital transitions. Regarding China’s practice, empirical research supports a dual regulatory logic of advancing digital application promotion and environmental performance standards in parallel. This provides a basis for mandating the integration of BIM and carbon management platforms in projects such as Xiongan New Area and the Guangdong-Hong Kong-Macao Greater Bay Area, and promotes the inclusion of environmental risk early warning into construction permit and acceptance systems. For the international community, particularly developing regions facing similar challenges in balancing infrastructure growth with sustainability, this study highlights the potential of a “digitalization + sustainability” synergy. It suggests that integrating digital tools with governance mechanisms can be a viable pathway to optimize procurement and project delivery, offering a comparative perspective for global stakeholders rather than a universally prescriptive framework.

Therefore, the proposed “institution-driven digital transformation” model can be adapted and implemented through the specific actions within the Chinese context, serving as a benchmark for others: Government departments should play a role in top-level design by establishing a standard system for the digital transformation of the construction industry (such as specifications for BIM technology application and evaluation standards for smart construction) and supporting incentive policies (such as preferential green credit and tax breaks) to provide institutional safeguards for technology implementation. Meanwhile, they should establish a cross-departmental collaborative regulatory platform to address the issue of data silos. As the primary implementers, enterprises must establish a data-model-decision closed-loop system. For example, enterprises should utilize IoT to monitor construction safety risks in real time, employ AI algorithms to optimize resource allocation, and incorporate green innovation efficiency into project performance metrics, thereby forming a technology-driven closed-loop risk management system. Industry associations should assume responsibility for capacity building by developing tiered and categorized training programs (such as certification for smart construction engineers) and establishing industry knowledge-sharing platforms. They should also lead the development of a digital transformation maturity model to provide benchmarking and diagnostic services for enterprises. Through a dynamic optimization mechanism involving policy piloting, enterprise feedback, and standard iteration, the three parties will ensure the integrated advancement of institutional frameworks and technological evolution.

7.2. Limitations and Prospects

The sample is concentrated on A-share listed companies in China, which imposes certain limitations on the generalizability of the conclusions. Future research should incorporate project data from multiple countries along the Belt and Road Initiative to conduct a comparative analysis of institutional contexts.

The environmental governance indicators rely on corporate disclosures, which may cause measurement biases. It is recommended to integrate objective data, such as satellite remote sensing and IoT sensors, to construct a multi-source validated indicator system.

The human resource allocation dimension does not elaborate on digital skill elements; future studies need to include dynamic human capital measurement tools for digital literacy and human–machine collaboration capabilities.

This study translates theoretical insights into industry action guidelines: the success or failure of digital transformation ultimately depends on the deep integration of technological and organizational management capabilities. Only by embedding tools into governance processes, converting data into decision-making basis, and allocating talent to critical nodes can the construction industry fortify its risk defenses in an era of uncertainty and achieve the dual goals of high-quality and sustainable development.