Investigating Subcontracting Partnership in Sustainable Urban Transportation System Design

Abstract

This study investigates the role of subcontracting partnerships in enhancing collaboration and sustainability in urban rail transit system design, addressing the challenges posed by fragmented practices and environmental goals under China’s “Dual Carbon” policy. Using a mixed-methods approach, we integrate structural equation modeling (SEM) and factor analysis to identify critical success factors (CSFs) and their impacts on design performance. SEM, a statistical technique capable of analyzing complex relationships between unobservable “latent variables” (e.g., trust, innovation) and measurable outcomes, was employed to validate the hypothesized relationships among five key factors: Excellence in Quality, Interactive Collaboration, Collaborative Vision, Risk Strategy, and Strategic Innovation. Factor analysis consolidated 19 CSFs from the literature into these five constructs, explaining 69.09% of the variance. The SEM results revealed that Excellence in Quality, Interactive Collaboration, Risk Strategy, and Strategic Innovation directly improve design performance, while Collaborative Vision indirectly influences outcomes through mediating effects on risk management and innovation. These findings provide actionable strategies for leveraging BIM/blockchain tools and institutional frameworks to enhance sustainability in urban transportation projects. By contextualizing partnership dynamics within China’s state-led infrastructure ecosystem, this research enriches the theoretical understanding of partnership mechanisms.

1. Introduction

With the acceleration of urbanization in China, urban rail transit systems have become critical infrastructure for ensuring efficient mobility, alleviating congestion, reducing environmental pollution, and supporting sustainable development [1]. Over the past decade, China’s urban rail transit networks have undergone unprecedented expansion. According to the latest data from the China Association of Metros, by 2023, the total operational mileage of urban rail transit in mainland China has exceeded 10,000 km, accounting for 40% of the global total [2]. However, this rapid growth has introduced systemic challenges including increasing technological complexity [3], cost overruns caused by fragmented subcontracting practices [4], and heightened requirements for lifecycle environmental impact under the national “Dual Carbon” goals [5].

Design-phase decisions fundamentally determine total project costs and environmental footprints [6], making cross-disciplinary collaboration a cornerstone of quality control. Yet empirical studies reveal that in China’s rail transit sector, subcontractors frequently encounter technical conflicts due to insufficient understanding of professional interfaces or failures in performance-based dynamic evaluation mechanisms [7]. Traditional contract dispute resolution mechanisms prioritize liability demarcation. Paradoxically, this approach intensifies adversarial relationships, erodes trust, and impedes continuous improvement [8].

Subcontracting partnerships have emerged as a crucial mechanism for addressing these challenges in the Chinese context. Practices from Beijing’s rail transit projects demonstrate that subcontracting partnerships can reduce design iteration costs through real-time information sharing [9], while achieving significant lifecycle carbon emission reductions compared to conventional approaches [10]. These implementations not only validate the effectiveness of subcontracting partnership models within China’s regulatory and technological ecosystem but also align with the strategic requirements for sustainable transportation systems outlined in the national “14th Five-Year Plan for Modern Integrated Transportation Systems Development” [11].

Despite significant progress, existing research predominantly focuses on owner–contractor partnerships [12], with insufficient attention to collaborative dynamics within design teams. This knowledge gap directly hinders the development of adaptive frameworks tailored to rapidly evolving technologies and culturally diverse teams. To address this, this study employs structural equation modeling (SEM) and factor analysis to identify critical success factors (CSFs) in design collaboration. We propose, for the first time, a dynamic governance framework that integrates collaboration efficiency with sustainability metrics. This framework not only bridges the theoretical void in quantifying multidisciplinary collaboration dynamics but also provides scalable decision-making tools for cross-cultural teams. It thereby offers theoretical insights for complex system governance and equips urban rail transit project management with actionable strategies for sustainable implementation.

2. Literature Review

2.1. The Role of Partnership in the Project Design

In the construction industry, parties collaborate to share risks and benefits, integrate technical expertise, and build more efficient supply chains [13]. Partnerships are widely regarded as the most critical factor influencing project performance, avoiding delays, legal disputes, and cost overruns [14,15]. Studies demonstrate that partnerships with high levels of trust among collaborators can enhance communication efficiency [16], shorten project timelines, reduce costs [17], improve construction quality [18], and increase client satisfaction [19]. Additionally, they effectively resolve industry conflicts [20], mitigate risks, stimulate participant engagement and innovation [21], and enhance the competitiveness of all parties [22].

Researchers and practitioners have proposed various Critical Success Factors (CSFs) to promote partnerships in construction, the collected reference CSFs, as shown in Table 1. Notably, in sustainable urban transportation system design, these factors must be further integrated with sustainability objectives, such as lifecycle carbon emission control and co-ordination of multi-system technical standards [23,24]. For instance, the “professional teams” and “adaptability to change” proposed by Black et al. (2000) [19] can be contextualized in rail transit projects as interdisciplinary design team collaboration and rapid iteration capabilities for emerging technologies like BIM.

Focusing on subcontracting collaboration during the design phase, Chan et al. (2004) [18] highlighted that establishing conflict resolution strategies, fostering a win-win mindset, monitoring collaboration processes, clarifying responsibilities, resource sharing, and subcontractor involvement are critical to success. This finding is particularly relevant to urban transportation system design. For example, in metro projects, unresolved technical standard conflicts between electromechanical subcontractors and rail design teams may lead to interface errors and rework, directly increasing carbon emissions and costs [25]. Similarly, the “attitudinal and behavioral CSF dimensions” proposed by Tang et al. (2006) [26] explain the need for value alignment between design teams and subcontractors on sustainability goals, such as energy-efficient material selection.

Table 1. Critical Success Factors (CSFs) for subcontracting partnerships in construction industry: a systematic literature review.

Success Factor	References
1. Effective Communication	[27,28]
2. Technical Expertise	[29,30]
3. Win-Win Attitude	[18,31]
4. Goal Alignment	[21,32]
5. Commitment to Quality	[30,33]
6. Mutual Trust	[26,34]
7. Continuous Improvement	[19,26]
8. Flexibility to Change	[19,27]
9. Good Cultural Fit	[35,36]
10. Questioning Attitude	[19,31]
11. Resource Sharing	[27,30]
12. Top Management Support	[37,38]
13. Clear Definition of Responsibilities	[18,28]
14. Generation of Innovative Ideas	[18,39]
15. Effective Problem Solving	[26,40]
16. Risk Sharing	[19,41]
17. Long-Term Perspective	[28,38]
18. Performance Measurement	[17,42]
19. Cost Reduction	[19,43]

Table 1. Critical Success Factors (CSFs) for subcontracting partnerships in construction industry: a systematic literature review.

Success Factor	References
1. Effective Communication	[27,28]
2. Technical Expertise	[29,30]
3. Win-Win Attitude	[18,31]
4. Goal Alignment	[21,32]
5. Commitment to Quality	[30,33]
6. Mutual Trust	[26,34]
7. Continuous Improvement	[19,26]
8. Flexibility to Change	[19,27]
9. Good Cultural Fit	[35,36]
10. Questioning Attitude	[19,31]
11. Resource Sharing	[27,30]
12. Top Management Support	[37,38]
13. Clear Definition of Responsibilities	[18,28]
14. Generation of Innovative Ideas	[18,39]
15. Effective Problem Solving	[26,40]
16. Risk Sharing	[19,41]
17. Long-Term Perspective	[28,38]
18. Performance Measurement	[17,42]
19. Cost Reduction	[19,43]

Regarding technology-enabled collaboration, Chen et al. (2012) [27] used structural equation modeling (SEM) to identify collaborative culture and resource sharing as core success drivers. In urban transportation projects, BIM-based platforms can detect multi-disciplinary design clashes (e.g., conflicts between rail and power supply layouts) in real time, optimizing designs to reduce construction waste and improve resource efficiency [24]. Doğan et al. (2016) [28] further validated that effective co-ordination and shared goals mitigate design–construction interface risks, which is critical for highly complex projects like integrated transportation hubs (e.g., multimodal interchange centers).

Recent studies have explored the role of emerging technologies in enhancing construction partnerships. For instance, He et al. (2023) [44] demonstrated that BIM-based platforms strengthen trust and resource sharing through real-time transparency, laying the foundation for interactive collaboration in urban transportation system design. Waqar et al. (2024) [45] proposed the potential of blockchain technology in contract management to optimize risk-sharing mechanisms, such as using smart contracts to automatically enforce sustainable procurement clauses (e.g., recycled material usage ratios), ensuring subcontractor compliance with environmental goals. These technological advancements underscore the evolving role of subcontracting partnerships in supporting sustainability within digital-era urban transportation system design.

2.2. Performance Assessment of Design Work

The evaluation of project success requires a multidimensional framework. The traditional “iron triangle” (time, cost, and quality) remains foundational [46,47]. However, as project complexity increases, researchers have progressively incorporated soft indicators such as stakeholder satisfaction [48,49] and collaboration efficiency [50]. In the domain of design performance, existing studies emphasize its alignment with project objectives [51]. For instance, architectural design necessitates balancing comfort, maintainability, and environmental sustainability [52], while urban transportation system design prioritizes subsystem interoperability and technical standardization [53].

This study positions sustainability as a core evaluation dimension, operationalized into three traceable hard metrics at the design stage: design quality (quality compliance documents), schedule adherence (milestone approval records), and cost efficiency (audit reports). While prior research has expanded sustainability criteria to include social benefits (e.g., public engagement) and lifecycle carbon emissions [54,55], such metrics rely on post-construction monitoring data and cannot be validated in real time during the design phase. Subcontracting partnerships resolve technical conflicts at the design stage (e.g., reconciling fire safety codes with material recyclability requirements), and their solutions are documented in design change files, which serve as direct evidence for quality compliance assessments [28]. Consequently, this study establishes an evaluation framework based on the aforementioned verifiable metrics, ensuring objectivity and practicality through their direct linkage to contractual documentation.

2.3. Evaluation Methods for Partnerships in the Construction Industry

With the expansion of global construction project scales and the increasing complexity of stakeholder relationships, the scientific evaluation of partnership quality has become a shared focus in academia and industry. In the assessment of partnerships in the construction industry, methods such as the Analytic Hierarchy Process (AHP), Fuzzy Comprehensive Evaluation (FCE), and Data Envelopment Analysis (DEA) have been widely adopted. For instance, Kim and Nguyen (2018) [56] quantified the construction supply chain (CSC) relationship framework using AHP, but its weight allocation heavily relied on subjective expert judgment, making it difficult to avoid human bias. Similarly, Noorizadeh (2021) [57] applied the DEA model to analyze supplier performance in construction supply chains, but this method could only handle quantitative indicators, overlooking soft factors such as differences in construction project environments that are difficult to directly observe. Multi-criteria decision frameworks (e.g., TOPSIS) can integrate multiple objectives but fail to explain causal relationships between variables (Zhang et al., 2024) [58]. The core issues with these methods lie in indicator redundancy, missing latent variables, and neglect of dynamic interactions. For example, Yang et al. (2024) [59] found that existing evaluation systems often include dozens of explicit indicators (e.g., government guarantees, social support), but the key factors actually influencing partnership performance may be hidden in a few latent dimensions. Malik et al. (2021) [60] further pointed out that variables like “communication” indirectly affect project outcomes through mediating paths such as “conflict”, yet traditional methods lack the capability to model such complex mechanisms.

To address the issue of indicator redundancy, factor analysis achieves dimensionality reduction by extracting common factors. For example, Faris et al. (2022) [61] conducted exploratory factor analysis (EFA) on 11 initial indicators in their study of partnerships in construction projects in emerging economies, ultimately identifying six core factors such as “project vision” and “participant behavior”, with a cumulative variance explanation rate of 45.9%. This process not only simplified the evaluation system but also revealed key dimensions hidden in the raw data. Adu Gyamfi et al. (2024) [62] performed factor analysis on 19 initial risk management indicators for public–private partnership (PPP) projects in Ghana’s construction industry, aggregating them into financial resource management, material resource management, and labor resource management, which together constituted the key elements of risk resource management (RRM). However, factor analysis only addresses indicator aggregation and cannot explain the pathways of influence between variables. The introduction of structural equation modeling (SEM) compensates for this limitation. SEM employs a dual framework of measurement models (observed variables → latent variables) and structural models (paths between latent variables), simultaneously enabling the quantification of latent variables and the testing of causal hypotheses. Empirical research by Sun et al. (2018) [63] demonstrated that factors in construction project co-operation networks and project characteristics can influence knowledge transfer carrier factors, thereby affecting the effectiveness of knowledge transfer—an indirect effect that traditional regression models cannot capture. Recent studies have further expanded the application scenarios of SEM. For instance, Almarri and Boussabaine (2025) [64] used SEM to identify key success factors and their inter-relationships in PPP projects for smart city infrastructure, exploring the intrinsic mechanisms through which these factors influence project success and providing a scientific basis for effective project implementation. These findings not only validate the statistical power of SEM but also highlight its unique advantages in uncovering complex mechanisms of influence.

The integration of factor analysis and structural equation modeling provides a systematic solution for evaluating construction partnerships. For example, in a study of PPP project partnerships in Vietnam, researchers first identified core factors such as “public sector cluster” and “private sector cluster” through factor analysis, then used SEM to examine the impact of multiple latent factors on the successful implementation of PPP infrastructure projects (Hai et al., 2022) [65]. In a study of urban rail transit PPP project partnerships in China, researchers first categorized factors influencing project success into five groups—personnel factors, policy factors, environmental factors, transaction structure factors, and project characteristic factors—using factor analysis, then employed SEM to demonstrate the specific pathways and inter-relationships of these factors on project success. This framework not only streamlines the evaluation process but also identifies key nodes for management intervention through path analysis (Liu et al., 2023) [66].

The evaluation of partnerships in the construction industry is shifting from “experience-driven” to “model-driven”. The integration of factor analysis and structural equation modeling not only addresses the shortcomings of traditional methods in latent variable extraction and causal inference but also provides a methodological foundation for building refined and dynamic evaluation systems. Given the content and objectives of research on urban rail transit design subcontracting partnerships, factor analysis and structural equation modeling undoubtedly offer unique rationality and advantages. Therefore, this paper adopts factor analysis and structural equation modeling (SEM) as the evaluation tools for the study.

3. Methodology

The methodology of this study consists of five steps, as illustrated in Figure 1.

Step 1: Through an extensive literature review, the widely-recognized Critical Success Factors (CSFs) in the construction industry are first identified. Then, these factors are adjusted according to the characteristics of urban rail transit design to narrow down the research scope.

Step 2: Starting from the CSFs summarized in the literature review, these factors specific to design subcontracting partnerships are refined and validated through questionnaire surveys, so as to determine the Success Factors (SFs) for design subcontracting partnerships in urban rail transit and the design performance indicators.

Step 3: By applying factor analysis, numerous original success factor variables identified in Step 2 are integrated into several representative common factors, which are regarded as the Critical Success Factors (CSFs) for subcontracting partnerships in urban rail transit design, thus facilitating more in-depth subsequent analysis. On this basis, a path model is proposed to hypothesize the logical connections among the critical success factors and between these factors and the overall design performance, visually presenting the potential relationships among variables and establishing a clear framework for empirical validation.

Step 4: Based on Step 3, a path model is proposed to hypothesize the logical relationships between critical success factors and their collective impact on overall design performance. This model visually presents potential variable interactions and establishes a clear framework for empirical validation.

Step 5: Utilize the Structural Equation Model (SEM) to verify the model established in Step 3 and the research hypotheses, so as to reveal the intrinsic mechanisms by which the critical success factors influence the design performance.

3.1. Data Collection Methods

This study adopts a mixed-method approach to collect data, ensure comprehensiveness and reliability.

3.1.1. Questionnaire Design

The questionnaire design meticulously integrates relationships among research variables, focusing on critical success factors (CSFs) and design performance-related variables. For instance, “effective communication” is measured through dimensions such as the openness of communication channels, accuracy of information exchange, and its impact on design decision-making, holistically evaluating communication’s transparency, effectiveness, and role in shaping decisions. “Resource sharing” is explicitly categorized into knowledge sharing and physical resource sharing to prevent respondent ambiguity. “Design quality” is assessed not only against contractual standards but also via a multidimensional professional scale encompassing innovation, functionality, safety, and other criteria, ensuring objective and comprehensive evaluation of outcomes.

3.1.2. Pilot Survey and Final Distribution

A pilot survey involving 30 designers from diverse regions and institutions was conducted to refine the questionnaire. Ambiguities in phrasing (e.g., “resource sharing”) were identified and revised to enhance clarity and reliability. The finalized questionnaire was distributed via stratified sampling (emails, phone calls, and in-person delivery) to 290 experts—managers, project leaders, and designers—from urban rail transit design institutes of varying scales and regions. Using a 1–5 Likert scale, the survey assessed performance metrics such as design quality, schedule adherence, cost control, and client satisfaction. A total of 223 valid responses were obtained, achieving an 86% valid response rate.

3.1.3. Regional Coverage

Survey regions were strategically selected to capture geographical and developmental diversity: first-tier cities (e.g., Beijing, Shanghai, Guangzhou) with mature rail transit systems, advanced technologies, and diverse partnership models; second-tier cities (e.g., Chengdu, Wuhan, Xi’an) undergoing rapid transit development, balancing challenges and opportunities; and cities across eastern, central, and western China to reflect regional contrasts. Sample distribution included 45 questionnaires in Beijing (42 valid), 42 in Shanghai (35 valid), 38 in Guangzhou (35 valid), and 155 in second-tier cities (111 valid). Analysis revealed regional disparities: first-tier cities excelled in technological innovation and resource sharing, while second-tier cities prioritized cost control and local adaptability, underscoring the regional impact on partnership dynamics.

3.1.4. Cultural Diversity Considerations

Cultural influences on partnerships were explored through the literature and field research, identifying key dimensions: values (divergent risk tolerance, innovation prioritization, and quality emphasis across regions), communication styles (direct vs. indirect approaches), and work attitudes (individual achievement vs. team collaboration). To capture these nuances, a customized cultural background section was added to the questionnaire, gathering data on respondents’ regional and educational cultural contexts. This enabled analysis of how cultural backgrounds shape collaborative behaviors and attitudes, offering empirical insights into cultural impacts on partnership effectiveness. Demographic information of respondents is detailed in

Table 2. Demographic profile of survey respondents in Chinese urban rail transit design partnerships (N = 223).

Category	Indicator Description	Number of Respondents	Percentage (%)
Age	20–30 years	55	24.6%
	31–40 years	77	34.5%
	41–50 years	68	30.5%
	Over 50 years	23	10.4%
Gender	Male	151	67.7%
	Female	72	32.3%
Position	Management of Design Institutes	81	36.3%
	Project Managers	69	30.9%
	Designers	73	32.8%
Education Level	High School/Technical Secondary School	0	0
	College	27	12.1%
	Bachelor	147	65.9%
	Graduate and Above	49	22.0%
Work Experience	0–5 years	38	17.0%
	6–10 years	76	34.1%
	11–20 years	68	30.5%
	Over 21 years	41	18.4%
Regional Cultural Background	Eastern coastal areas	58	26.0%
	Central regions	60	26.9%
	Western regions	53	23.8%
	Northeast regions	52	23.3%
Educational Cultural Background	Mainland China	173	77.6%
	Hong Kong, Macao, and Taiwan regions of China	23	10.3%
	European and American countries	22	9.9%
	Other countries	5	2.2%

.

3.2. Data Analysis Methods

3.2.1. Basics of Factor Analysis Methods

Factor analysis is a statistical method used to simplify data structures by extracting a few comprehensive factors from numerous original variables, thereby explaining the variability in the data more concisely [67]. First, the Kaiser–Meyer–Olkin (KMO) test and Bartlett’s test of sphericity are conducted to assess the suitability of the data for factor analysis. The KMO measure of sampling adequacy is calculated as follows:

$$KMO={\displaystyle \frac{{\sum }_{i\ne j}^p{r}_{ij}^2}{{\sum }_{i\ne j}^p{r}_{ij}^2+{\sum }_{i\ne j}^p{u}_{ij}^2}}$$

where $r_{ij}$ is the correlation coefficient between variables i and j, and $u_{ij}$ is the partial correlation coefficient between variables i and j. A KMO value closer to 1 indicates stronger correlations among variables, supporting the use of factor analysis. Bartlett’s test of sphericity tests the null hypothesis that the correlation matrix is an identity matrix. The test statistic is ${\mathrm{X}}^2$, which is calculated based on the determinant of the correlation matrix ($\left|\mathbf{R}\right|$) and the sample size (n):

$${\mathrm{X}}^2=-[\mathrm{n}-1-{\displaystyle \frac{2p+5}{6}}]\ln \left|\mathbf{R}\right|$$

where p is the number of variables. A significant Bartlett’s test result (with a p-value < 0.05) further validates the appropriateness of factor analysis. If the data meets these criteria, Kaiser-normalized Varimax rotation is applied to the factor loading matrix. The Varimax rotation aims to maximize the sum of the variances of the squared factor loadings for each factor, which enhances the interpretability of the factor structure. Mathematically, for a factor loading matrix A with elements $a_{ij}$, the Varimax criterion is defined as:

$$\mathrm{V}={\displaystyle \frac{1}{m}}{\sum }_{j=1}^m\left[{\sum }_{i=1}^p{\left(a_{ij}^2-\overline{a_j^2}\right)}^2\right]$$

where m is the number of factors, p is the number of variables, and $\overline{a_j^2}={\displaystyle \frac{1}{p}}{\sum }_{i=1}^p{a}_{ij}^2$. Through this process, the numerous success factors influencing design and subcontracting partnerships are consolidated into key factors, which serve as critical variables in subsequent analyses to clarify the mechanisms by which complex partnerships affect design performance.

3.2.2. Hypothetical Path Model Construction

Constructing the hypothetical path model involves establishing logical relationships between key success factors (CSFs) and overall design performance based on factor analysis results. After preprocessing data to ensure accuracy and address missing values, validated factors derived from factor analysis are integrated into the model under the assumption of positive correlations with design performance, consistent with theoretical and practical insights in urban rail transit design. The model further hypothesizes interactions among key factors, such as synergistic enhancements between factors, reflecting their interdependence in collaborative partnerships. This framework enables empirical validation of relationships through statistical methods like structural equation modeling (SEM), ultimately clarifying how CSFs collectively influence design outcomes.

3.2.3. Structural Equation Modeling (SEM) Analysis

The hypothetical path model is tested using Amos 26 and Maximum Likelihood Estimation to analyze questionnaire data. Structural equation modeling is a comprehensive statistical approach that can simultaneously evaluate multiple independent and dependent variables, account for latent variables and measurement errors, and examine complex inter-relationships. The general form of a structural equation model can be written as two sets of equations: measurement model and structural model [68].

• Measurement model analysis

The measurement model represents the relationships between latent variables (ξ and η) and their observed indicators (x and y). For example, for the exogenous latent variables ξ:

$$x_i={\lambda }_{x_i}{\xi }_i+{\delta }_i$$

where $x_i$ is the i-th observed indicator of the exogenous latent variable ξ, ${\lambda }_{x_i}$ is the factor loading representing the relationship between the latent variable and the observed indicator, and δ_i is the measurement error.

For the endogenous latent variables η:

$$y_j={\lambda }_{y_j}{\eta }_j+{\epsilon }_j$$

where $y_j$ is the j-th observed indicator of the endogenous latent variable η, ${\lambda }_{y_j}$ is the factor loading, and ${\epsilon }_j$ is the measurement error.

In this study, Amos 26 was used to conduct reliability and validity tests to ensure the robustness of the measurement. A Cronbach’s alpha coefficient greater than 0.7 indicates good internal consistency. Convergent validity was verified through standardized factor loadings (≥0.6, p < 0.001), composite reliability (CR > 0.7), and average variance extracted (AVE > 0.5). Discriminant validity was evaluated by comparing the square root of the average variance extracted with the correlations between variables to ensure the differences between latent structures. These steps guaranteed the reliability and validity of the data and enhanced the credibility of the subsequent analysis results.

• Structural model analysis

The structural model represents the relationships among latent variables:

$$\mathsf{\eta }=\mathrm{B}\mathsf{\eta }+\mathsf{\Gamma }\mathsf{\xi }+\mathsf{\zeta }$$

where B is a matrix of coefficients representing the relationships among endogenous latent variables, Γ is a matrix of coefficients representing the relationships between exogenous and endogenous latent variables, and ζ is the disturbance term.

Model fit indices are assessed through multiple criteria. For example, the χ²/degrees of freedom (χ²/df) ratio is used to evaluate the overall fit of the model. A value between 1 and 2 is generally considered acceptable. The Goodness-of-Fit Index (GFI) is calculated as:

$$\mathrm{G}\mathrm{F}\mathrm{I}=1-{\displaystyle \frac{{\sum }_{i=1}^n{(y_i-\widehat{y_i})}^2}{{\sum }_{i=1}^n{(y_i-\overline{y})}^2}}$$

where $y_i$ is the observed value, $\widehat{{\mathrm{y}}_i}$ is the predicted value, and $\overline{\mathrm{y}}$ is the mean of the observed values. A GFI value closer to 1 indicates a better fit.

The Adjusted Goodness-of-Fit Index (AGFI) adjusts the GFI for the degrees of freedom:

$$\mathrm{A}\mathrm{G}\mathrm{F}\mathrm{I}=1-(1-\mathrm{GFI}){\displaystyle \frac{n-1}{n-k-1}}$$

where n is the number of observations and k is the number of estimated parameters.

The Root Mean Square Error of Approximation (RMSEA) is another important index:

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}\mathrm{A}=\sqrt{{\displaystyle \frac{{\chi }^2-df}{df(n-1)}}}$$

An RMSEA value of less than 0.05 indicates a very good fit, and values between 0.05 and 0.1 are considered at the threshold.

The Comparative Fit Index (CFI), Tucker–Lewis Index (TLI), Normed Fit Index (NFI), Incremental Fit Index (IFI), and Relative Fit Index (RFI) also contribute to evaluating the model fit. For example, the CFI is calculated based on the non-centrality parameters of the null and alternative models:

$$\mathrm{C}\mathrm{F}\mathrm{I}=1-{\displaystyle \frac{\max (0,{\chi }_0^2-d{f}_0)}{\max (0,{\chi }_1^2-d{f}_1)}}$$

where ${\mathsf{\chi }}_0^2$ and $\mathrm{d}{\mathrm{f}}_0$ are the ${\mathsf{\chi }}^2$ value and degrees of freedom of the null model, and ${\mathsf{\chi }}_1^2$ and $\mathrm{d}{\mathrm{f}}_1$ are those of the alternative model.

Path coefficients are analyzed for significance to quantify the impact of CSFs on design performance. A significant path coefficient indicates a meaningful relationship between the corresponding variables, providing robust support for research conclusions.

4. Results

4.1. Survey Results

This study aggregated and averaged the success factor ratings provided by all respondents, with the specific results presented in

Table 3. Measured variables of Success Factors (SFs) and design performance metrics: survey rating scores (SV = Success Variable; M = Performance Metric).

Category	Factor	Mean Value	Standard Deviation
Partnership Variables	Effective Communication (SV1)	3.79	1.019
	Technical Expertise (SV2)	3.71	1.139
	Win-Win Attitude (SV3)	3.50	1.048
	Goal Alignment (SV4)	3.69	0.995
	Commitment to Quality (SV5)	3.54	1.025
	Mutual Trust (SV6)	3.65	1.075
	Continuous Improvement (SV7)	3.77	1.165
	Flexibility to Change (SV8)	3.63	1.086
	Good Cultural Fit (SV9)	3.80	1.130
	Questioning Attitude (SV10)	3.82	0.985
	Resource Sharing (SV11)	3.77	1.118
	Top Management Support (SV12)	4.00	1.038
	Clear Definition of Responsibilities (SV13)	3.52	1.056
	Generation of Innovative Ideas (SV14)	3.91	0.933
	Effective Problem Solving (SV15)	3.70	1.101
	Risk Sharing (SV16)	3.66	1.056
	Long-Term Perspective (SV17)	3.86	1.017
	Performance Measurement (SV18)	3.53	1.122
	Cost Reduction (SV19)	4.04	0.967
Design Performance	Design Quality (M1)	3.87	1.003
	Design Schedule (M2)	3.93	1.050
	Cost-Effectiveness (M3)	3.96	0.990

. The table reveals that the mean scores of all success factors exceed 3.5, surpassing the average score of 3. This demonstrates that all critical success factors for subcontracting partnerships have been effectively incorporated in urban rail transit system design, and they collectively exhibit positive contributions to enhancing design performance.

Among the numerous success factors, the average score of “Effective Communication (SV1)” reached 3.79, with a standard deviation of 1.019. These data imply that, overall, the communication channels between design teams and subcontractors are relatively smooth, information transfer is relatively accurate, and it has a positive impact on design decision-making. However, the relatively large standard deviation also suggests that there are significant differences in communication effectiveness among different projects or teams.

The average score of “Technical Expertise (SV2)” is 3.71, with a standard deviation of 1.139. This reflects that the technical capabilities of relevant personnel are widely recognized. However, due to the large standard deviation, it indicates that there are differences in the levels of technical expertise among different projects or teams, which may be affected by factors such as project scale, complexity, and regional technological development levels.

“Top Management Support (SV12)” received a high score of 4.00, with a standard deviation of 1.038, fully highlighting the crucial role of top management in promoting subcontracting partnerships.

In terms of design performance indicators, the average score of “Design Quality (M1)” is 3.87 (standard deviation = 1.003), the average score of “Design Schedule (M2)” is 3.93 (standard deviation = 1.050), and the average score of “Cost-Effectiveness (M3)” is 3.96 (standard deviation = 0.990). These data show that relatively good performance has been achieved in these aspects. However, the standard deviations also indicate that there is still room for improvement in design quality, schedule, and cost-effectiveness in urban rail transit system design projects. These results lay the foundation for the subsequent discussion of the impact of various factors on design performance and also echo the direction of enhancing project performance in the research conclusions.

Based on the above data, Figure 2 visually presents the scores of success factors and design performance indicators in the form of a bar chart. From the figure, it is clearer to compare the differences in the scores of various factors. For example, “Top Management Support (SV12)” and “Cost Reduction (SV19)” have relatively high scores among the success factors, highlighting the importance of these two factors in the project. In terms of design performance indicators, the score of “Cost-Effectiveness (M3)” is relatively high, while the score of “Design Quality (M1)” is slightly lower, visually demonstrating the performance differences in design performance in different aspects. Through this bar chart, we can more intuitively grasp the data characteristics, providing convenience for further analysis.

4.2. Factor Analysis

Based on the results of the questionnaire survey, this study conducted the Kaiser–Meyer–Olkin (KMO) test and Bartlett’s test of sphericity. The KMO value is 0.857, which indicates a strong correlation among the variables, making it suitable for factor analysis. In addition, the significance level of Bartlett’s test of sphericity (Sig. < 0.001) further confirms the correlation among the variables, supporting the decision to carry out factor analysis. The “Kaiser-normalized Varimax rotation” was applied to the factor loading matrix, and this method converged after six iterations. The analysis resulted in five factors. Figure 3 is the scree plot of the factor analysis, where the abscissa represents the factor number, and the ordinate represents the eigenvalue. It can also be clearly seen from the figure that these five factors have a strong explanatory power for data variation.

Table 4. Variance explained by principal components in factor analysis of design partnership success factors (KMO = 0.857, Cumulative Variance = 69.09%).

Factor	Initial Eigenvalues			Rotation Sum of Squared Loadings
	Total	Variance Percentage	Cumulative %	Total	Variance Percentage	Cumulative %
F1	5.710	30.053	30.053	3.366	17.714	17.714
F2	2.667	14.037	44.090	2.762	14.538	32.253
F3	1.866	9.821	53.911	2.712	14.271	46.524
F4	1.638	8.620	62.531	2.193	11.544	58.068
F5	1.246	6.559	69.090	2.094	11.022	69.090

shows that these five factors account for 69.090% of the total variance, collectively capturing most of the variation in the success factor variables (SVs). They effectively represent the original success factor variables and provide a set of more concise and interpretable variables for subsequent analysis.

Based on the initial eigenvalues and the rotation sum of squared loadings, all CSFs have been grouped into five eigenvectors. Therefore, this study named all the vectors based on their characteristics as Excellence Quality Factor (F1), Interactive Collaboration Factor (F2), Collaborative Vision Factor (F3), Risk Strategy Factor (F4), and Strategic Innovation Factor (F5).

Excellence Quality Factor: The main factor F1 has loadings greater than 0.5 on the indicators of Commitment to Quality (SV5), Questioning Attitude (SV10), Effective Problem Solving (SV15), Performance Measurement (SV18), and Continuous Improvement (SV7). It covers the unwavering commitment to quality, encourages questioning of existing processes to promote innovation, emphasizes the ability to solve problems effectively, and monitors and improves outcomes through performance measurement. Additionally, this factor emphasizes the importance of continuous improvement, i.e., constantly seeking opportunities to improve efficiency and effectiveness to achieve excellent management practices.

Interactive Collaboration Factor: The main factor F2 has loadings greater than 0.5 on the indicators of Effective Communication (SV1), Mutual Trust (SV6), Resource Sharing (SV11), and Good Cultural Fit (SV9). These elements together promote interactive collaboration between design teams and subcontracting units, establishing a close co-operative relationship. Through open and honest communication, both parties can build trust, share resources, and strengthen teamwork through cultural fit, all of which are key to achieving common goals.

Collaborative Vision Factor: The main factor F3 has loadings greater than 0.5 on the indicators of Goal Alignment (SV4), Flexibility to Change (SV8), Win-Win Attitude (SV3), and Long-Term Perspective (SV17). These elements ensure that design teams and subcontracting units work toward the same direction, can adapt to changes flexibly, seek win-win solutions, and consider the sustainability of co-operation from a long-term perspective. This shared vision helps partners stay united and motivated in the face of challenges.

Risk Strategy Factor: The main factor F4 has loadings greater than 0.5 on the indicators of Clear Definition of Responsibilities (SV13), Risk Sharing (SV16), and Cost Reduction (SV19). These elements help ensure that design teams and subcontracting units can jointly bear responsibilities, reasonably allocate risks, and take measures to reduce costs when facing potential challenges and uncertainties. This proactive risk management helps reduce the risk of project delays and cost overruns, improving the overall financial health of the project.

Strategic Innovation Factor: The main factor F5 has loadings greater than 0.5 on the indicators of Top Management Support (SV12), Technical Expertise (SV2), and Generation of Innovative Ideas (SV14). These indicators highlight the central role of these factors in promoting the strategic development of partnering. Support from top management provides direction and resources for partnering, technical expertise ensures the efficiency and quality of execution, and the generation of innovative ideas is the source of continuous competitive advantage. These elements work together to promote the strategic implementation of partnering and stimulate innovative vitality, allowing partnering to maintain a leading position in the market.

Table 5. Rotated factor loadings of Critical Success Factors (CSFs) using varimax rotation (loadings > 0.75 highlighted in bold).

Variables	Principal Factor
	F1	F2	F3	F4	F5
SV1	0.251	0.764	0.261	0.025	0.152
SV2	0.156	0.203	0.226	0.115	0.762
SV3	0.087	0.048	0.754	0.205	0.132
SV4	−0.016	0.166	0.781	0.134	−0.026
SV5	0.779	0.233	0.079	0.139	−0.016
SV6	0.215	0.737	0.151	0.063	0.030
SV7	0.794	0.057	−0.037	0.093	0.033
SV8	0.048	0.052	0.807	0.084	0.103
SV9	0.156	0.786	0.080	0.017	0.135
SV10	0.793	0.136	0.081	0.027	0.203
SV11	0.090	0.821	0.029	0.126	0.083
SV12	0.127	0.059	−0.048	0.056	0.811
SV13	0.092	0.057	0.234	0.820	−0.001
SV14	0.029	0.093	0.110	−0.013	0.839
SV15	0.820	0.147	−0.011	0.031	0.091
SV16	0.102	0.028	0.250	0.773	0.074
SV17	0.102	0.246	0.777	0.285	0.079
SV18	0.771	0.173	0.146	0.091	0.081
SV19	0.107	0.116	0.114	0.837	0.070

lists the factor loadings of all eigenvectors (principal factors). The bold values are greater than 0.75.

4.3. Hypothetical Path Model

Based on the principal factors identified in the factor analysis, a hypothetical path model (as shown in Figure 4) was proposed to represent the logical connections between the CSFs and overall design performance. The hypothetical path model includes several influencing paths that need to be examined.

The factors of Excellence in Quality (F1), Interactive Collaboration (F2), Collaborative Vision (F3), Risk Strategy (F4), and Strategic Innovation (SI) are positively correlated with design performance.

As the Collaborative Vision (F3) improves, the factors of Interactive Collaboration (F2), Risk Strategy (F4), and Strategic Innovation (F5) become stronger.

As the Interactive Collaboration (F2) improves, the factors of Excellence in Quality (F1) and Strategic Innovation (F5) become stronger.

4.4. SEM Analysis

4.4.1. Measurement Model Analysis

The reliability and convergent validity of the scale measuring the impact of partnering on design performance were tested using Amos 26 software. The results are shown in

Table 6. Reliability and validity metrics for extracted partnership factors (α = Cronbach’s Alpha; CR = Composite Reliability; AVE = Average Variance Extracted).

Indicator	Loading	α	CR	AVE
F1		0.872	0.874	0.581
SV5	0.780
SV7	0.714
SV10	0.771
SV15	0.781
SV18	0.762
F2		0.834	0.836	0.561
SV1	0.838
SV6	0.692
SV9	0.730
SV11	0.728
F3		0.834	0.881	0.651
SV3	0.710
SV4	0.699
SV8	0.679
SV17	0.884
F4		0.801	0.803	0.577
SV13	0.802
SV16	0.722
SV19	0.752
F5		0.770	0.770	0.532
SV2	0.858
SV12	0.616
SV14	0.693
M		0.856	0.857	0.666
M1	0.846
M2	0.808
M3	0.794

. Specifically, “Loading” refers to the standardized factor loading, α is Cronbach’s alpha, “CR” is composite reliability, “AVE” is average variance extracted, and the observed variables are the simplified expressions.

The reliability test results show that the Cronbach’s alpha coefficients for the Excellence in Quality Factor (F1), Interactive Collaboration Factor (F2), Collaborative Vision Factor (F3), Risk Strategy Factor (F4), Strategic Innovation Factor (F5), and Design Performance (M) constructs are all greater than 0.7. These results indicate good internal consistency of the sample data. From the results of the confirmatory factor analysis, it is evident that the standardized factor loadings of all observed indicators for the constructs of Excellence in Quality (F1), Interactive Collaboration (F2), Collaborative Vision (F3), Risk Strategy (F4), and Strategic Innovation (F5) with the construct of Design Performance (M) are above 0.6, and all are significant at the p < 0.001 level. The composite reliability (CR) is greater than 0.7, and the AVE values are greater than 0.5, indicating that the overall measurement model has a good fit. To verify the discriminant validity among the constructs, a correlation analysis was conducted, and the square root of the AVE was calculated, as shown in

Table 7. Discriminant validity analysis: correlations between latent constructs vs. square roots of AVE (bold diagonal values).

Factors	F1	F2	F3	F4	F5	M
F1	0.762
F2	0.413 *	0.749
F3	0.191 *	0.333 *	0.747
F4	0.242 *	0.217 *	0.456 *	0.760
F5	0.251 *	0.311 *	0.249 *	0.176 *	0.729
M	0.443 *	0.488 *	0.285 *	0.376 *	0.354 *	0.816

Note: * indicates significance at the p < 0.05 level, and the parts in bold black are the square roots of AVE.

. The square root of the AVE is greater than the correlation coefficients between each construct, indicating that the measurement model has good discriminant validity.

4.4.2. Structural Model Analysis

In the structural equation modeling (SEM) analysis, the model fit indices serve as the core basis for judging the degree of fit between the hypothesized path model and the actual data. In this study, multiple fit indices were used to comprehensively evaluate the model, and the results are shown in

Table 8. Structural Equation Model (SEM) goodness-of-fit indices for hypothesized partnership–performance relationships.

Goodness-of-Fit Measures	Recommended Level of GOF Measure	Initial Statistical Values (11-Path Model)	Final Statistical Values (10-Path Model)
${\mathsf{\chi }}^2/\mathrm{degree}\mathrm{of}\mathrm{freedom}(\raisebox{1ex}{${\chi }^2$}\!\left/ \!\raisebox{-1ex}{$df$}\right.$)	1 to 2	1.186	1.190
Goodness-of-fit index (GFI)	0 (no fit)–1 (perfect fit)	0.911	0.910
Adjusted Goodness-of-fit index (AGFI)		0.887	0.887
Root mean sq. error of approx (RMSEA)	<0.05 (very good)–0.1 (threshold)	0.029	0.029
Comparative fit index (CFI)	0 (no fit)–1 (perfect fit)	0.983	0.982
Tucker–Lewis index (TLI)	0 (no fit)–1 (perfect fit)	0.980	0.980
Normal fit index (NFI)	0 (no fit)–1 (perfect fit)	0.901	0.900
Incremental fit index (IFI)	0 (no fit)–1 (perfect fit)	0.983	0.983
Relative fit index (RFI)	0 (no fit)–1 (perfect fit)	0.885	0.885

. The initial value of the chi-square to degrees of freedom ratio ($\raisebox{1ex}{${\chi }^2$}\!\left/ \!\raisebox{-1ex}{$df$}\right.$) is 1.186, and the final value is 1.190, which falls within the ideal range of 1–2, indicating that the overall model fit is good and can effectively explain the data variation. The initial value of the Goodness-of-Fit Index (GFI) is 0.911, and the final value is 0.910. The Adjusted Goodness-of-Fit Index (AGFI) is 0.887. All these values are close to 1, further verifying the adaptability of the model to the data. The Root Mean Square Error of Approximation (RMSEA) is 0.029, which is less than 0.05, showing an excellent fit between the model and the data. In addition, the Comparative Fit Index (CFI), Tucker–Lewis Index (TLI), Normed Fit Index (NFI), Incremental Fit Index (IFI), and Relative Fit Index (RFI) are all at relatively high levels, fully demonstrating that the model can better reflect the relationships among variables.

In Figure 5, the horizontal axis shows the various fit indices of the structural equation model, and the vertical axis represents the index values. The columns of different colors represent the 11-path model and the 10-path model, respectively. By comparing the lengths of different columns, it can be directly observed that the two models perform similarly in terms of data fitting ability, adaptability to data, and the ability to reflect variable relationships, and they have similar reliability and validity.

The analysis of standardized path coefficients is crucial for quantifying the impact of critical success factors (CSFs) on design performance.

Table 9. Standardized path coefficient estimates in Structural Equation Modeling (SEM) analysis of subcontracting partnership impacts on sustainable design performance (model with 11 paths).

Path	Standardized Coefficient Estimate	S.E.	t-Value	Sig.
F2←F3	0.53	0.092	5.718	***
F4←F3	0.65	0.097	6.727	***
F5←F3	0.28	0.114	2.457	*
F1←F2	0.47	0.074	6.423	***
F5←F2	0.37	0.101	3.665	***
M←F2	0.32	0.092	3.520	**
M←F4	0.32	0.087	3.677	***
M←F1	0.25	0.081	3.122	**
M←F5	0.18	0.067	2.690	**
M←F3	−0.14	0.106	−1.339

Note: * indicates significance at p < 0.05, ** indicates significance at p < 0.01, and *** indicates significance at p < 0.001.

presents the relevant data under the 11-path model. The standardized path coefficient of Interactive Collaboration (F2) on Design Performance (M) is 0.32 (p < 0.01), indicating a significant positive impact on design performance. That is, good interactive collaboration can effectively improve design performance. The coefficient of Risk Strategy (F4) is 0.32 (p < 0.001), Excellence in Quality (F1) is 0.25 (p < 0.01), and Strategic Innovation (F5) is 0.18 (p < 0.05), all of which have a positive effect on design performance. However, the coefficient of Collaborative Vision (F3) on design performance is −0.14 (p > 0.05), indicating no direct significant impact, which is different from the traditional partnership theory.

Table 10. Standardized path coefficient estimates in Structural Equation Modeling (SEM) analysis of subcontracting partnership impacts on sustainable design performance (model with 10 paths).

Path	Standardized Coefficient Estimate	S.E.	t-Value	Sig.
F2←F3	0.52	0.092	5.671	***
F4←F3	0.65	0.097	6.670	***
F5←F3	0.27	0.113	2.413	*
F1←F2	0.47	0.074	6.421	***
F5←F2	0.37	0.101	3.700	***
M←F2	0.28	0.087	3.231	**
M←F4	0.25	0.069	3.640	***
M←F1	0.26	0.081	3.260	**
M←F5	0.16	0.066	2.495	*

Note: * indicates significance at p < 0.05, ** indicates significance at p < 0.01, and *** indicates significance at p < 0.001.

shows the situation of the 10-path model. The standardized path coefficient of Interactive Collaboration (F2) on Design Performance (M) is 0.28 (p < 0.01), Risk Strategy (F4) is 0.25 (p < 0.001), Excellence in Quality (F1) is 0.26 (p < 0.01), and Strategic Innovation (F5) is 0.16 (p < 0.05). The trend is consistent with that in Table 9, reconfirming the positive impact of these factors on design performance. The impact of Collaborative Vision (F3) on design performance remains insignificant, further indicating that in practical situations, due to the lack of an institutionalized mechanism for transforming shared goals into actionable indicators, it is difficult to directly promote the improvement of design performance.

Figure 6 confirms and modifies the hypothesized path model in Figure 3 based on the standardized path coefficient data from Table 10. Specifically, Interactive Collaboration (F2), Risk Strategy (F4), Excellence in Quality (F1), and Strategic Innovation (F5) have significant positive impacts on Design Performance (M), with standardized path coefficients of 0.28 (p < 0.01), 0.25 (p < 0.001), 0.26 (p < 0.01), and 0.16 (p < 0.05), respectively, effectively enhancing design performance; for example, efficient Interactive Collaboration (F2) can improve communication and co-operation between design teams and subcontractors, promoting design project progress. Notably, although Collaborative Vision (F3) has no direct significant impact on design performance (the coefficient is −0.14, p > 0.05), it plays an indirect role by influencing F2, F4, and F5. A clear and shared Collaborative Vision (F3) can boost team communication and co-ordination, which benefits Interactive Collaboration (F2), helps to formulate more scientific Risk Strategy (F4), and provides a direction for Strategic Innovation (F5), offering support for the overall progress of design projects. Figure 6 visually presents the relationships among these factors, offering a basis for understanding the role of subcontracting partnerships in design.

5. Discussion

5.1. Interpretation of Key Findings

The SEM results validate the hypothesized relationships among the five identified factors, offering nuanced insights into the dynamics of subcontracting partnerships in sustainable urban transportation design. Interactive Collaboration (F2) emerged as the strongest direct predictor of design performance (β = 0.28, p < 0.01), aligning with prior studies emphasizing communication and trust as foundational to cross-disciplinary co-ordination in complex infrastructure projects [27]. This finding underscores the critical role of real-time information sharing and conflict resolution mechanisms—such as BIM-based platforms—in mitigating design-phase inefficiencies. For instance, the β value for F2’s impact on Excellence in Quality (F1) (β = 0.47, p < 0.001) suggests that collaborative workflows directly improve technical outcomes, corroborating evidence from BIM implementation in Shenzhen Metro Line 9 [69].

Strategic Innovation (SI) exerted a significant positive effect on design performance (β = 0.16, p < 0.05), reflecting the transformative potential of top management support and technological integration. This aligns with Waqar et al.’s (2024) [45] findings on blockchain’s role in enhancing transparency and sustainability compliance. Conversely, Collaborative Vision (CV) demonstrated no direct impact on performance (β = −0.14, p > 0.05), a result that contradicts traditional partnership theory [19]. This discrepancy may stem from the operationalization of vision in China’s context, where shared goals often lack institutionalized mechanisms for translation into actionable metrics. As North (1990) [70] posits, formalized procedures are critical for converting vision into practice, which may be absent in state-led projects prioritizing speed over cultural alignment.

5.2. Practical Implications for Urban Transportation Projects

The validated success factors provide actionable strategies to address systemic challenges in China’s rail transit industry. Interactive Collaboration (IC) can be institutionalized through BIM-based co-design platforms, as similar studies have shown that such platforms can reduce rework costs. For example, the Shenzhen Metro Line 9 project significantly improved design efficiency by resolving numerous electromechanical–architectural conflicts during the design phase through cloud-based clash detection [69]. Strategic Innovation (SI) can be embedded in lifecycle sustainability frameworks, leveraging blockchain-enabled smart contracts for design optimization [71]. For projects with weak Collaborative Vision (CV) effects, managers should prioritize cultural brokerage mechanisms and short-cycle feedback loops to reconcile differences in risk tolerance and innovation priorities. For instance, studies have shown that project co-ordinators in transnational projects successfully reduced communication risks, ensuring smooth project execution [72].

5.3. Theoretical Contributions

The non-significant direct path between Collaborative Vision (CV) and design performance expands partnership theory by introducing contextual mediation thresholds. While shared goals remain a necessary condition, their sufficiency depends on three boundary conditions: institutionalization depth, cultural cohesion index, and temporal flexibility. Institutionalization depth requires formalized procedures (e.g., ISO 19650 standards [73]) to translate vision into actionable standards; the cultural cohesion index emphasizes the alignment of team values, which can be quantified using tools such as Hofstede’s cultural dimensions; and temporal flexibility refers to the feasibility of iterative adjustment periods, often constrained by China’s accelerated project schedules. This theoretical reconfiguration bridges the gap between Western partnership models [69] and China’s state-led, efficiency-driven infrastructure ecosystem. For example, Western partnership theories emphasize the role of legal frameworks (e.g., NEC4 contracts) in safeguarding vision, whereas in China’s state-led projects, informal relational governance may weaken the immediacy of vision realization, necessitating cultural brokerage mechanisms and short-cycle feedback loops.

5.4. Limitations

This study has several limitations: The empirical data were primarily collected from Chinese urban rail transit projects (77.6% of respondents educated in mainland China’s cultural context, Table 2), and the findings may be constrained by China’s centralized decision-making culture (e.g., the high score of “Top Management Support” at 4.00, Table 3) and the stratified regional sampling (Section 3.1.3), limiting their generalizability to regions with decentralized governance or distinct cultural norms. Additionally, the cross-sectional survey design restricts causal inferences about long-term impacts, as the sustainability of factors like the “Strategic Innovation Factor” (F5) with high loadings on technical expertise (0.762) and innovative ideas (0.839) (Table 5) under evolving technological landscapes remains unverified.

6. Conclusions

This research employs a mixed-methods approach to explore subcontracting partnerships, addressing the critical challenge of enhancing collaboration in the sustainable design of urban transportation systems. At the core of the analysis lies structural equation modeling (SEM), a statistical technique capable of analyzing complex relationships between unobservable “latent variables” (e.g., trust, innovation) and measurable outcomes. By integrating SEM with factor analysis, five key drivers of successful partnerships were systematically identified: Excellence in Quality (EQ), Interactive Collaboration (IC), Collaborative Vision (CV), Risk Strategy (RS), and Strategic Innovation (SI), and their direct and indirect impacts on design performance were quantified. The findings reveal that, except for Collaborative Vision (CV), the other four factors have the most significant direct effects on sustainability outcomes. Collaborative Vision does not directly improve performance but acts as a “relational amplifier,” enhancing risk management and collaboration when supported by institutional frameworks (e.g., standardized procedures) and cultural alignment. These insights were validated using SEM, which accounts for measurement errors and complex interdependencies, ensuring robust results.

The study provides multifaceted practical implications for policymakers, industry leaders, and researchers. For policymakers and industry leaders, it is essential to actively leverage digital tools, such as adopting BIM platforms for real-time conflict resolution and utilizing blockchain technology to automate sustainability compliance processes. Additionally, aligning contracts with standards like ISO 19650 can transform vague “collaborative visions” into actionable metrics. In transnational projects, cultural awareness should be prioritized by appointing project co-ordinators to effectively bridge regional and cultural differences. For researchers, the proposed SEM framework offers a replicable tool for studying partnerships in complex infrastructure systems, effectively balancing technical and human factors, and further emphasizes the necessity of constructing context-specific governance models in diverse cultural and regulatory environments. These insights provide theoretical support and practical guidance for improving collaboration efficiency in sustainable urban transportation system design.

Although the findings advance partnership theory, these conclusions are primarily based on China’s state-led infrastructure ecosystem. Therefore, future research should validate the applicability of the SEM framework in regions with different governance models (e.g., Europe, North America), explore long-term sustainability impacts through longitudinal studies, and develop tools to quantify “cultural cohesion” and “institutionalization depth,” thereby enhancing the broad applicability and practical value of the research outcomes.