Modeling the Multiple Driving Mechanisms and Dynamic Evolution of Urban Inefficient Land Redevelopment: An Integrated SEM-FCM Approach

Abstract

Urban inefficient land redevelopment (UILR) is crucial for sustainable urban development, yet its progress is driven by the interplay of multiple factors. To systematically uncover the driving mechanisms and dynamic patterns of these factors, an integrated analytical approach combining Structural Equation Modeling (SEM) and Fuzzy Cognitive Map (FCM) is developed in this study. Based on 222 valid survey responses from professionals across eight cities in China’s Yangtze River Delta region, five key factors are identified within the “drivers–pressure–enablers” conceptual framework: economic incentives, environmental objectives, social needs, policy guidance, and implementation conditions. SEM is first employed to examine static causal relationships, and the quantified pathway effects are subsequently incorporated into an FCM model to simulate the long-term evolution. The results reveal the following: (i) All five factors exert significant direct effects, with economic incentives, environmental objectives, and policy guidance also demonstrating notable indirect effects. (ii) The factors exhibit distinct temporal characteristics: policy guidance acts as a “fast variable” enabling short-term breakthroughs; economic incentives serve as a “strong variable” driving medium-term progress; and social needs function as a “slow variable” with long-term benefits. (iii) Policy guidance is essential, as its absence leads to persistently low effectiveness, while its synergy with implementation conditions can achieve satisfactory performance even without economic incentives. The combined SEM–FCM approach validates static hypotheses and simulates dynamic scenarios, offering a new perspective for analyzing complex driving mechanisms of UILR and providing practical insights for targeted redevelopment strategy design.

1. Introduction

As urbanization in China transitions from a phase of rapid growth to one of stable development, the urban development model is undergoing a profound transformation from extensive expansion to connotative enhancement [1]. Extensive utilization of land resources, the core carrier of urban space, has become a key bottleneck constraining high-quality urban development [2]. A large amount of land is utilized inefficiently, including old industrial zones, urban villages, and abandoned sites, triggering a series of challenges in urban areas, such as environmental degradation, social segregation, and declining economic vitality [3]. In this context, urban inefficient land redevelopment (UILR) has emerged as an essential strategic choice for revitalizing the existing urban stock, optimizing resource allocation, and promoting sustainable urban development [4].

However, UILR constitutes a complex systemic process involving multiple stakeholders, including governments, developers, property owners, and the public, and encompasses multiple objectives such as economic returns, social equity, and environmental sustainability [5]. In practice, conflicts among diverse interests, as well as trade-offs between competing goals, often lead to significant implementation barriers, slow progress, and difficulties in achieving a balance in the overall benefits. Although existing studies have identified various driving factors from policy, market, and social angles, most analyses remain static and fragmented. How these factors interact to form driving mechanisms, and how such mechanisms evolve under changing conditions, remains a “black box”, limiting theoretical support for systematic intervention and informed regulation [6].

To address these gaps, this study places the multiple driving factors of UILR within a unified analytical framework, adopting a hybrid approach that integrates structural association and dynamic evolution to enable a more comprehensive analysis of UILR’s driving mechanisms. Drawing on Yang’s triple-structure theory of development dynamics [7], the driving factors are deconstructed into three core dimensions: drivers, pressures, and enablers. Structural Equation Modeling (SEM) [8] is first used to empirically examine the structural relationships among these factors and identify key pathways and effect strengths. Subsequently, the core variables and causal relationships derived from SEM are translated into a Fuzzy Cognitive Map (FCM) [9] as conceptual nodes and weighted connections, forming a dynamic cognitive model for scenario simulation. The structural complementarity between SEM and FCM bridges static validation with dynamic simulation, providing robust methodological support for understanding the driving mechanisms of UILR.

The contributions of this study are threefold. Theoretically, it introduces the “drivers–pressures–enablers” framework in UILR research, providing a new perspective for understanding complex interactions among multiple factors. Methodologically, by integrating SEM and FCM, this study achieves both the static verification and dynamic simulation of driving mechanisms, deepening the understanding of UILR as a complex adaptive system. Practically, the findings provide urban decision-makers with a decision-support tool for scenario simulation and policy experimentation, aiding in the formulation of more forward-looking, adaptive, and targeted redevelopment strategies to promote high-quality and sustainable urban development.

2. Literature Review

2.1. Multiple Stakeholders in UILR

Urban Inefficient Land (UIL) refers to developed urban parcels that are underused, vacant, derelict, or functionally obsolete, failing to realize their full economic, social, or environmental potential [10]. The redevelopment of such land, known as UILR, serves as a critical strategy for advancing urban intensification and sustainable development. Consequently, UILR has emerged as a key research focus in the fields of urban planning, land management, and public policy [11]. Accordingly, systematically identifying stakeholders and analyzing their behavioral motivations are essential for addressing implementation barriers and refining policies [12].

In stakeholder research, the analytical focus has evolved beyond the traditional “government–market” dichotomy to embrace a tripartite “government–market–society” analytical framework [13,14]. Scholarly consensus highlights their distinct roles: local governments act as regulators and development promoters, influenced by performance metrics, fiscal pressures, and land finance mechanisms [15]; market entities, especially developers, are driven by economic returns and assess risks versus rewards [16]; while social actors, including residents and community organizations, show rising rights awareness and participation, with demands extending beyond compensation to include environmental, cultural, and social sustainability, which makes their involvement essential to redevelopment [17].

Methodologically, research has shifted from analyzing static roles to dynamic interactions, with recent work emphasizing stakeholders’ strategic interdependencies across redevelopment phases [18]. For instance, Han et al. applied evolutionary game theory to model intensive land use and multi-actor decisions [5], while Cao et al. detailed government-resident negotiations, collective action in property integration, and tripartite bargaining over benefits [19]. Building on traditional game-theoretic approaches, current research has widely incorporated complex-system simulation methods such as social network analysis [20,21] and stakeholder value network [2], along with mixed-method tools including multi-agent simulation [22], and spatial econometric models [23]. These multidimensional analyses reveal the interactions and strategic mechanisms among actors across stages, providing theoretical and empirical support for collaborative governance and sustainable redevelopment models [24].

2.2. Multiple Driving Factors and Mechanisms of UILR

UILR is propelled by the interplay of multiple factors. Existing research primarily focuses on two major themes: the identification of driving factors [25] and their underlying mechanisms [4]. Regarding driving factors, a systematic understanding has emerged. Economic factors form the fundamental endogenous impetus for redevelopment [26]. Policy factors provide the legal foundation and institutional framework for redevelopment projects [27]. Environmental and social factors jointly constitute important external conditions [28,29].

Following the identification of driving factors, research has shifted to examine their interactive mechanisms and implementation pathways. Analytical frameworks have moved from linear models to complex-system interaction [30]. For instance, Li et al. demonstrated through game theory that negotiations among government, developers, and property owners directly shape project outcomes [11]. Fan et al. argued that negotiation, community engagement, and benefit-sharing can integrate stakeholders, resolve conflicts, and support sustainable redevelopment [31]. Recently, influenced by social-governance theory, studies have increasingly emphasized tripartite collaboration among government, market, and society [3].

Yet existing research remains largely focused on material-spatial determinants [32] and often overlooks redevelopment’s core role in value creation and balancing diverse interests [33]. Due to socioeconomic and institutional variations across cities, deeper mechanisms like policy implementation, market response, and community participation remain underexamined in specific contexts [34]. Furthermore, as re-development increasingly focuses on existing urban stock, practical challenges such as benefit allocation, property-rights integration, and land assembly have become pressing concerns for policy and research [35].

To address these complex challenges, research has increasingly focused on the dynamic evolution of urban renewal systems [36], aiming to capture their nonlinear, time-varying characteristics shaped by multi-factor interactions [37]. Grounded in complex adaptive systems theory [38], current work examines how economic, social, and environmental subsystems co-evolve through adaptive behavior. Dynamic simulation methods—notably system dynamics [39] and multi-agent systems [20]—are widely used to model the temporal evolution of key variables like urban resilience and land use structure under policy interventions and external shocks.

2.3. SEM and FCM Methods

Structural equation modeling (SEM) combines confirmatory factor analysis and path analysis to quantify causal relationships and compare their relative importance [8]. In fields like urban renewal, SEM is commonly used to identify causal pathways among factors and evaluate the strength of these relationships [40]. Fuzzy Cognitive Mapping (FCM), on the other hand, simulates the dynamic behavior of systems through conceptual nodes and causal weights, making it suitable for modeling fuzzy associations and complex feedback systems. It has been applied in research areas including urban sustainability and resilience assessment [41], and collaborative governance in old neighborhood renovation [42]. However, the determination of causal weights in FCM often relies on expert judgment, which introduces subjective limitations [43].

In recent years, hybrid SEM–FCM approaches have attracted growing attention. This methodology first applies SEM to identify key variables and estimate path coefficients from empirical data, which then inform the initial weights in an FCM. FCM is subsequently used for dynamic simulation to examine system behavior and feedback under various scenarios [44,45]. Combining SEM’s empirical rigor with FCM’s dynamic simulation capacity, the integrated approach has been effectively applied in areas such as construction safety risk [9], project complexity management [46], corporate social responsibility [47], and international reputation mechanisms [48]. The SEM-FCM hybrid method thus offers an integrated research tool for analyzing the driving mechanisms of complex systems such as UILR, effectively bridging static validation with dynamic simulation.

In summary, research on UILR has shifted from linear analysis to a complex systems perspective, emphasizing dynamic interactions and collaborative governance among stakeholders in areas such as property-rights integration, benefit distribution, and spatial reproduction. Methods including evolutionary game theory [49], agent-based modeling [36], social network analysis [22], and system dynamics [50] are widely used to uncover causal mechanisms and identify sustainable pathways within specific institutional contexts. However, three gaps persist: (i) the absence of a systematic theoretical integration of multiple drivers means that their internal structure is still unknown; (ii) most mechanism studies remain conceptual and lack quantitative tests, leaving related hypotheses unverified; and (iii) the dynamic evolution of the driving system is under-explored, preventing simulations of how external shocks affect system responses and steady states. To address these gaps, in this study, multiple driving factors are incorporated within a unified “drivers–pressures–enablers” analytical framework, and SEM is integrated into FCM to develop a comprehensive methodology combining static structural analysis with dynamic scenario simulation.

3. Methodology

In this study, an integrated analytical approach combining SEM and FCM is developed to systematically examine the driving mechanisms underlying UILR. SEM is employed to analyze the static structural relationships among driving factors, assessing latent variables and measurement errors to quantify causal pathways and effect strengths, and thereby revealing the system’s inherent logical structure. FCM complements this approach by simulating dynamic evolution processes, projecting transition pathways from current states to future steady states and forecasting system responses under various policy scenarios.

3.1. SEM Framework Specification

SEM represents a comprehensive multivariate statistical technique that integrates factor and path analyses. This methodology is particularly suitable for addressing complex research problems involving multiple indicators and variables and enables quantitative assessment of abstract theoretical constructs while effectively minimizing subjective bias, thereby enhancing both the objectivity and practical interpretability of analytical outcomes. In this study, SEM is employed to examine the static causal relationships among different driving factors, thereby revealing the underlying mechanisms of UILR. SEM was selected primarily because it allows for the statistical validation of all factors within a complete hypothesized model, ensuring the consistency of the theoretical framework. Furthermore, the developed SEM will serve as an empirical foundation for constructing a subsequent FCM model.

The SEM framework comprises two distinct components: the measurement model and the structural model. The measurement model specifies the relationships between observed variables and their underlying latent constructs, formally expressed through the following Equations [51]:

$$X=\lambda x\xi +\delta$$

(1)

$$Y=\lambda y\eta +\epsilon$$

(2)

In Equation (1), X denotes the vector of exogenous observed variables, ξ represents the exogenous latent variables, λx constitutes the factor loading matrix, defining the relationships between exogenous observed and latent variables, and δ refers to the measurement error terms associated with exogenous variables.

In Equation (2), Y represents the endogenous observed variables, η denotes the endogenous latent variables, λy constitutes the factor loading matrix, defining the relationships between endogenous observed and latent variables, and ε refers to the measurement error terms associated with endogenous variables.

The structural model characterizes the causal relationships among latent variables, formally expressed as follows:

$$\eta =\gamma \xi +\beta \eta +\zeta$$

(3)

In Equation (3), both γ and β denote path coefficients, with β specifically representing the relationships between endogenous latent variables, and γ capturing the directional influences of exogenous latent variables on endogenous latent variables. The term ζ signifies the residual error term of the structural equation.

3.2. SEM Validation

SEM validation proceeds in two stages: measurement model assessment followed by structural model assessment. First, confirmatory factor analysis (CFA) is conducted to examine the reliability and validity of the measurement model. This involves calculating standardized factor loadings for observed variables, along with composite reliability (CR) and average variance extracted (AVE) for latent variables, to assess whether the internal consistency, convergent validity, and discriminant validity meet established psychometric standards. Subsequently, multiple goodness-of-fit indices (GOF), including the χ²/degrees of freedom ratio (χ²/df), goodness-of-fit index (GFI), comparative fit index (CFI), Tucker–Lewis index (TLI), Standardized Root Mean Square Residual (SRMR) and root mean square error of approximation (RMSEA), are used to comprehensively evaluate the alignment between the theoretical model and empirical data [42]. Following the model fit assessment, maximum likelihood estimation is applied to derive standardized path coefficients. These coefficients are then statistically examined using critical ratios and significance levels to test the research hypotheses: paths demonstrating statistical significance (using p < 0.05 as the conventional threshold) support hypothesis acceptance, whereas non-significant paths warrant retention of the null hypothesis. The final output is a structured hypothesis test report that includes path coefficients (β), standard errors (S.E.), critical ratios (C.R.), and significance levels (P), thereby providing quantitative evidence to support the research conclusions.

3.3. FCM Model Construction

Complex systems generally comprise multiple interconnected elements that form dynamic networks through nonlinear relationships [33]. FCM, as a soft computing technique integrating fuzzy logic and neural network principles, is an effective approach to simulate the dynamic behavior of such systems and conduct decision-oriented scenario analyses [52]. This methodology utilizes directed graph structures to represent causal relationships among conceptual nodes within a system, thereby enabling the transformation of qualitative cognitive mappings into semi-quantitative dynamic models. This capability allows researchers to simulate system evolution trajectories and steady-state characteristics under varying conditions.

In urban systems, FCM has been applied in areas such as analyzing the driving factors in old industrial building renovation, assessing the sustainability of large urban infrastructure [53], evaluating safety risks in utility tunnel operation and maintenance [9], and examining value co-destruction in urban village renovation [40]. These applications illustrate the method’s ability to model complex causal relationships and incorporate stakeholder perceptions in loosely structured urban environments, thereby supporting its relevance for UILR analysis as proposed in this study.

To explore the dynamic evolution of driving mechanisms, an FCM model is developed based on the SEM. As a dynamic modeling tool, FCM simulates system behavior through weighted causal relationships among conceptual nodes. The path coefficients validated by SEM reflect the degree to which driving factors influence redevelopment performance. Therefore, the validated and statistically significant standardized path coefficients from SEM serve as initial weights for the FCM. This approach addresses the limitations of traditional FCM, which relies on subjective weight assignment, while providing an empirically grounded starting point for dynamic simulation [46]. Accordingly, based on empirical findings from SEM analysis, an FCM framework is constructed to specifically simulate dynamic mechanisms through which various driving factors influence the performance of UILR.

The FCM model characterizes the strength of the association between nodes Ci and Cj as a “weight”, denoted by Wij, which takes a value between −1 and 1. The absolute value of Wij reflects the strength of the causal link: a larger absolute value indicates a stronger causal relationship. Specifically, a negative weight (Wij < 0) implies that an increase in Ci inhibits Cj, leading to a decrease in its value; a positive weight (Wij > 0) means that an increase in Ci facilitates Cj, resulting in an increase in its value; and a zero weight (Wij = 0) indicates the absence of any causal connection between the nodes.

Formally, a standard FCM structure can be represented by a quadruple U = (C, W, A, f ), where C = [C1,C2,⋯,Cm] denotes the set of m conceptual nodes, W refers to an m × n weight matrix that quantifies the strength and direction of causal relationships between nodes [54]; A = Ai (t) represents the time-varying state value of node Ci at time step t; and f is a monotonically increasing nonlinear threshold function that aggregates the inputs from all nodes influencing the target node into its activation domain. The state value of node Cj at the next time step t + 1 is computed as follows:

$$A_j(t+1)=f\left({\displaystyle \sum _{\begin{array}{c}i=1\\ \end{array}}^m{A}_i(t)W_{ij}}\right)$$

(4)

Commonly used activation functions include the sigmoid function and the hyperbolic tangent function. The sigmoid function constrains node states to the interval [0, 1], while the hyperbolic tangent function bounds node states within [−1, 1]. Compared with the sigmoid function, the hyperbolic tangent aligns more closely with real-world causal logic, where effects can be either promoting (positive) or inhibiting (negative), while also enhancing the interpretability of concept-state changes. In addition, the steeper gradient of the hyperbolic tangent near the origin accelerates model convergence during dynamic iterations. For these reasons, the hyperbolic tangent function (shown in Equation (5)) is employed in this study to more clearly distinguish between positive and negative correlations.

$$f(x)={\displaystyle \frac{1-e^{-2ax}}{1+e^{-2ax}}}$$

(5)

In this equation, the parameter *a* > 0 controls the steepness of the threshold function.

3.4. FCM Simulation Analysis

FCM-based simulation encompasses three primary analytical modes: predictive, diagnostic, and hybrid analysis. Predictive analysis involves projecting a system’s output state and evolutionary trajectory from a predefined initial state, while diagnostic analysis, in contrast, infers the necessary adjustments in causal nodes by specifying the desired states of target nodes. Hybrid analysis integrates both approaches: it first employs predictive analysis to reveal the system’s natural evolution path, then applies diagnostic analysis to identify input modifications required to achieve specific objectives, thereby establishing a closed-loop decision-support framework [55].

4. Model Establishment

4.1. Identification of Driving Factors

4.1.1. Classification of Driving Factors

Regarding the forces that drive development, Chinese scholar Yang introduced a framework known as the “triple dynamics theory of development” or the “triple-structure theory of developmental dynamics” in his philosophical work Research on the Handling and Resolution of Contradictions [7]. This theory has since been applied in studies on dynamic and driving mechanisms [56].

Based on Yang’s triple-structure theory of development dynamics, the forces propelling system evolution can be classified into three categories: drivers, pressures, and enablers [57]. Drivers constitute intrinsic and proactive motivations originating within the system, representing the fundamental source of developmental momentum. Pressures refer to external influences that compel systemic adaptation, typically manifested as resource constraints or external challenges. Enablers encompass supportive elements that facilitate system evolution by creating favorable conditions and removing implementation barriers. This framework provides a logically coherent and structurally organized theoretical lens through which to examine the driving factors in UILR. Taking the redevelopment of underutilized industrial land as an example, developers’ expectations of land-value appreciation and the government’s need to increase fiscal revenue constitute the drivers; soil-remediation regulations, ecological redlines and carbon-emission standards create environmental pressure, while residents’ demands for public space and a better urban image generate social governance pressure [24]. At the same time, preferential land prices, tax reductions, and streamlined approval processes provide institutional enablers at the policy level, while special funds, mature technologies, and property rights compensation agreements constitute conditional enablers at the implementation level. Ultimately, the interaction of drivers, pressures, and enablers propels the renewal, transformation, and upgrading of the industrial area [28].

In this study, the driving factors of UILR are systematically categorized within the “drivers–pressures–enablers” framework as follows: drivers correspond to economic incentives, capturing the fundamental motivation of market entities and governments to obtain land premiums and fiscal revenues. Pressures include both environmental objectives and social needs, where the former denotes natural system pressures from land shortages and environmental pollution, while the latter represents societal pressures arising from public expectations for improved living conditions and urban functionality. Enablers encompass the institutional environment established through policy guidance, along with implementation conditions such as financial resources, technical capacity, and property rights integration. These three categories interact synergistically to form a driving system for UILR, with detailed classifications presented in

Table 1. Categorization of driving factors.

No.	Driving Factor	Category	Explanation
1	Economic incentives	Drivers	The pursuit of higher economic returns by market entities constitutes the fundamental driving force.
2	Environmental objectives	Pressure	Environmental constraints and sustainability objectives generate compelling pressure for redevelopment.
3	Social needs	Pressure	Public demands for improved living conditions and urban quality generate sustained social pressure.
4	Policy guidance	Enablers	Government policies and regulations provide directional guidance and institutional support.
5	Implementation conditions	Enablers	Financial resources, technical expertise, and property rights integration serve as essential enabling conditions.

.

4.1.2. Identification of Relevant Variables

A systematic literature review was conducted to identify variables relevant to the driving mechanisms of UILR, following a four-stage procedure. First, comprehensive searches of the China National Knowledge Infrastructure (CNKI) and Web of Science databases were performed using key terms including “urban land redevelopment,” “inefficient-land redevelopment,” and “urban renewal.” The search was restricted to publications from 2000 to 2025 in Chinese or English, indexed in CSSCI, CSCD, SSCI, or SCI, yielding 903 initial records. Second, screening of titles, abstracts, and keywords identified 111 publications closely aligned with the research focus. Third, specialized attention was given to studies addressing driving factors, mechanisms, or pathways, with backward reference tracking employed to enhance variable system comprehensiveness and reliability, resulting in 61 core publications. Finally, through systematic analysis of this literature corpus, six key factors and twenty corresponding measurement variables were identified, as shown in

Table 2. Factors associated with the driving mechanisms of UILR.

Variable	Measurement Items	Reference Source
Economic incentives (ECO)	Expected GDP growth rate (ECO1)	[39]
	Forecasted growth in property values (ECO2)	[4]
	Industrial upgrading and restructuring (ECO3)	[3]
	Expected local tax revenue increase (ECO4)	[20]
Environmental objectives (ENV)	Land resource Shortage (ENV1)	[36]
	Environmental pollution and ecological stress (ENV2)	[38]
	Infrastructure carrying capacity pressure (ENV3)	[34]
Social needs (SOC)	Demand for improved living environment (SOC1)	[25]
	Need for enhanced public service facilities (SOC2)	[5]
	Public support and engagement (SOC3)	[58]
Policy guidance (POL)	Level of support from higher-level policies (POL1)	[42]
	Compliance with urban planning (POL2)	[4]
	Guidance from specialized plans (POL3)	[29]
	Efficiency of administrative approval (POL4)	[59]
Implementation conditions (IMP)	Adequacy of funding (IMP1)	[8]
	Technical and plan feasibility (IMP2)	[31]
	Effectiveness of stakeholder coordination mechanisms (IMP3)	[22]
Redevelopment performance (RED)	Improvement in intensive land use level (RED1)	[21]
	Level of regional economic value-added (RED2)	[11]
	Overall satisfaction of surrounding residents (RED3)	[32]

.

4.2. Research Hypotheses and Path Model

Urban land is a rare capital resource; the fundamental impetus for UILR comes from urban land’s latent economic value. Stakeholders engage in redevelopment to profit from land appreciation, operational revenues, fiscal growth, and industrial upgrades. Empirical evidence consistently identifies economic feasibility as a critical determinant of redevelopment success [44]. Based on this insight, the following hypothesis is proposed:

Hypothesis 1 (H1).

Economic incentives positively influence redevelopment performance.

Resource limitations and environmental pressures function as external drivers of urban intensification. Constraints arising from land shortages and ecological carrying capacities compel city policymakers to prioritize stock revitalization, creating a compelling mechanism that integrates energy conservation, environmental protection, and livability objectives into redevelopment projects to achieve superior outcomes. Accordingly, the following hypothesis is proposed:

Hypothesis 2 (H2).

Environmental objectives positively influence redevelopment performance.

Social legitimacy represents a crucial foundation for redevelopment initiatives. Residents’ demands for better housing, expanded public spaces, and improved services require developers to respond effectively. Projects that secure community support and tangibly enhance residents’ well-being typically face fewer implementation barriers and lead to higher social satisfaction and overall performance. This supports the following hypothesis:

Hypothesis 3 (H3).

Social needs positively influence redevelopment performance.

Governmental intervention plays an instrumental role in coordinating resource allocation. Through regulatory frameworks, spatial planning, fiscal instruments, and policy directives, authorities establish institutional environments that guide market behavior toward alignment with broader urban development objectives. This leads to the following hypothesis:

Hypothesis 4 (H4).

Policy guidance positively influences redevelopment performance.

Implementation capacity provides the operational foundation for project realization. Financial resources, technical capabilities, administrative efficiency, and well-defined property rights collectively determine project advancement efficiency and output quality, forming essential prerequisites for achieving redevelopment targets [60]. This leads to the following hypothesis:

Hypothesis 5 (H5).

Implementation conditions positively influence redevelopment performance.

Policy functions as a potent tool for shaping market expectations. When a government designates an area as a priority renewal zone, the clear and encouraging signal boosts anticipated returns and invigorates investor enthusiasm. Complementary incentives such as tax rebates and additional floor area ratio bonuses directly improve the profitability of a project, thereby generating and amplifying economic gain motives. Based on this understanding, the following hypothesis is proposed:

Hypothesis 6 (H6).

Policy guidance positively influences economic incentives.

Governments, through the enactment of specific regulations, plans, and policies, can directly establish, facilitate, and secure the diverse conditions necessary for project implementation [46]. Specific measures include financial instruments to attract private investment, specialized agencies to enhance coordination, and standardized procedures to improve institutional safeguards. This supports the following hypothesis:

Hypothesis 7 (H7).

Policy guidance positively influences implementation conditions.

Environmental imperatives reconfigure implementation requirements and standards. Addressing ecological concerns demands advanced technical solutions, dedicated environmental funding, and specialized regulatory frameworks, transforming sustainability pressures into specific operational criteria [47]. Based on this insight, the following hypothesis is proposed:

Hypothesis 8 (H8).

Environmental objectives positively influence implementation conditions.

The pursuit of economic benefits often promotes high-intensity development through increased floor area ratios and energy-intensive industries. Such approaches accelerate resource consumption and increase pollutant emissions. Moreover, large-scale redevelopment frequently reveals underlying environmental issues like soil contamination, leading to the following hypothesis:

Hypothesis 9 (H9).

Economic incentives positively influence environmental objectives.

Positive financial projections motivate regulatory streamlining, infrastructure investment, and private sector participation, creating operational conditions conducive to project success. This leads to the following hypothesis:

Hypothesis 10 (H10).

Economic incentives positively influence implementation conditions.

The complete hypothesized model is presented in Figure 1.

4.3. Data Collection

A questionnaire survey was administered for data collection. Prior to full-scale distribution, a pilot study was conducted with seven experts with professional experience in redevelopment projects to validate the following: (i) the comprehensiveness of identified variables; (ii) the rationality of the classification framework; (iii) the plausibility of the hypothesized path model; and (iv) the clarity of questionnaire items. The pilot results confirmed that the variable system, classification approach, and path configuration were appropriate for UILR research. One expert observed that while “redevelopment performance” was conceptually defined as significant improvements in land use efficiency and urban function, the corresponding questions failed to adequately reflect this definition. Consequently, the relevant items were revised to better align with the definition. Additional explanatory notes were also added for selected variables to prevent misinterpretation.

After incorporating expert feedback, the final questionnaire comprised two sections: Section 1, in which respondents’ professional background information (

Table 3. Participants’ information.

Variable Type	Items	Options	Valid N	Percentage (%)
Basic information	Type of employer	Government department	14	6.30%
		Academia/Research institute	55	24.80%
		Development company	60	27.00%
		Community organization	48	21.60%
		Other	45	20.30%
	Years of experience	1–3 years	40	18.00%
		3–5 years	99	44.60%
		5–10 years	56	25.20%
		Over 10 years	27	12.20%
Professional relevance	Familiar with related work?	Very familiar	56	25.20%
		Fairly familiar	64	28.80%
		Moderately familiar	74	33.30%
		Not very familiar	28	12.60%
	Involved in related projects?	Yes	168	75.70%
		No	54	24.30%
Total			222	100%

) was collected to verify their domain expertise, and Section 2, in which a 5-point Likert scale was employed for each statement for respondents to indicate their agreement level. (See the Supplementary Materials for complete questionnaire content).

The survey was administered online and targeted professionals engaged in territorial spatial planning, urban renewal, land management and related policy research within eight representative cities of the Yangtze River Delta urban agglomeration: Shanghai, Nanjing, Wuxi, Changzhou, Suzhou, Hangzhou, Ningbo and Wenzhou. These cities were drawn from the official list of 43 inefficient land redevelopment pilot cities designated by China’s Ministry of Natural Resources (Ministry of Natural Resources (2023) No. 171, https://www.gov.cn/zhengce/zhengceku/202309/content_6903884.htm, accessed on 20 September 2025). The selected cities exemplify typical and leading characteristics in China’s urbanization process. They demonstrate not only high-density agglomeration and rapid development, but also face prominent challenges such as tight land-resource constraints and urgent spatial-restructuring needs. Therefore, they represent a highly relevant context for studying UILR.

Questionnaires were distributed through professional networks, industry forums and snowball sampling. A total of 242 responses were collected; after excluding submissions that were too short or exhibited obvious response patterns, 222 valid questionnaires remained, yielding an effective response rate of 91.7%. The questionnaire comprised 20 measurement items, yielding a sample-to-item ratio of 11.1:1, which exceeded the commonly recommended minimum of 10:1 and met the basic sample size requirement for SEM. All observed variables showed acceptable skewness and kurtosis values (|skewness| < 0.7, |kurtosis| < 1.2), satisfying the assumption of univariate normality. Mardia’s tests of multivariate skewness and kurtosis were performed using the MVN package in R via the R extension interface in SPSS 25.0. The results indicated no significant departure from multivariate normality (p > 0.05). The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was 0.835, exceeding the recommended threshold of 0.80. Bartlett’s test of sphericity was significant (χ² = 3422.03, df = 120, p < 0.001), indicating that the correlation matrix was suitable for factor analysis and structural equation modeling.

Given that the data were collected via self-report questionnaires, potential common method bias was addressed. Procedural remedies—including respondent anonymity, scale diversity, and the use of validated instruments—were implemented to mitigate its influence. Harman’s single-factor test was also conducted for statistical validation. Results from an unrotated exploratory factor analysis in SPSS 25.0 showed that the first factor explained 32.461% of the total variance, below the 50% threshold, indicating that common method bias did not substantially affect the findings.

4.4. Reliability and Validity Assessment

The measurement model was evaluated for reliability and validity using SPSS 25.0, with the following established thresholds: Cronbach’s α ≥ 0.7 for internal consistency; factor loading ≥ 0.6 for item reliability; average variance extracted (AVE) ≥ 0.5 for convergent validity; and composite reliability (CR) ≥ 0.70 for construct reliability.

As presented in

Table 4. Measurement model results.

Factor	Variable	Factor Loading	Mean	CR	AVE	Cronbach’s α
Economic incentives	C11	0.806	3.43	0.889	0.666	0.888
	C12	0.817	3.42
	C13	0.787	3.35
	C14	0.854	3.30
Environmental objectives	C21	0.777	3.39	0.856	0.664	0.854
	C22	0.816	3.36
	C23	0.850	3.41
Social needs	C31	0.807	3.54	0.862	0.676	0.862
	C32	0.829	3.62
	C33	0.830	3.66
Policy guidance	C41	0.820	3.57	0.871	0.629	0.871
	C42	0.763	3.62
	C43	0.784	3.61
	C44	0.804	3.73
Implementation conditions	C51	0.888	3.58	0.887	0.723	0.886
	C52	0.822	3.51
	C53	0.840	3.49
Redevelopment performance	CT1	0.852	3.35	0.879	0.707	0.878
	CT2	0.825	3.41
	CT3	0.846	3.38

, all constructs demonstrated values above the required limits, indicating that the sample data satisfy all measurement criteria and exhibit adequate reliability and convergent validity.

Goodness-of-fit indices were used to assess the model’s alignment with the empirical data. The results indicate an excellent fit: χ²/df = 1.347, GFI = 0.917, CFI = 0.978, TLI = 0.973, SRMR = 0.035 and RMSEA = 0.040. All values meet or exceed their respective acceptance thresholds, confirming strong support for the hypothesized factor structure.

The discriminant validity was evaluated to examine the degree of differentiation among the latent constructs in the model. The established criterion requires that the square root of the average variance extracted (AVE) for each construct be greater than its correlations with all other constructs. As shown in

Table 5. Discriminant validity.

Variable	Economic Incentives	Environmental Objectives	Social Needs	Policy Guidance	Implementation Conditions	Redevelopment Performance
Economic incentives	0.816
Environmental objectives	0.166	0.815
Social needs	0.247	0.211	0.822
Policy guidance	0.285	0.224	0.405	0.793
Implementation conditions	0.376	0.354	0.302	0.372	0.850
Redevelopment performance	0.442	0.367	0.403	0.459	0.481	0.841

, all six constructs meet this criterion, demonstrating adequate discriminant validity for the measurement model used in this study.

4.5. Hypothesis Testing and Path Analysis

Maximum likelihood estimation was chosen for its efficiency and asymptotic unbiasedness under normality assumptions. To confirm the suitability of this estimation approach, a Monte Carlo power analysis was first conducted for the target RMSEA = 0.05, α = 0.05, and df = 120. With 224 observations, the empirical power to reject the null hypothesis of close fit reached 0.85, exceeding the conventional threshold of 0.80 and thereby supporting the appropriateness of ML estimation. Subsequently, the ten hypotheses in the conceptual model were tested using maximum likelihood estimation in SEM. The initial results showed that Hypothesis 9, which posits that economic incentives positively influence environmental objectives, was not statistically significant (p > 0.05). This non-significant relationship may be explained by its dual nature: while economic expansion may intensify environmental pressures through resource consumption and pollution emissions, the resulting economic benefits can also provide financial support for adopting environmental technologies and promoting industrial upgrades, thereby potentially mitigating the ecological impacts. These countervailing “intensification” and “mitigation” effects operate simultaneously, leading to a non-significant net direct relationship in statistical terms. Due to its statistical non-significance, this path was excluded from the final model to improve model parsimony and precision. The removal of Hypothesis 9 indicates that environmental objectives function primarily as external constraints in the empirical model, not as endogenously driven mediators directly shaped by economic incentives. Within this framework, environmental dimensions are therefore better understood as independent governance objectives rather than as intermediary channels for transmitting economic effects. All subsequent analytical results and fit indices refer to this refined model (Figure 2).

Based on the fitting results of the revised model (as shown in

Table 6. Fitting results of the modified model.

N	Hypothesis	β	S.E.	C.R.	P	Sig.
H1	Economic incentives→ Redevelopment performance	0.196	0.079	2.332	***	support
H2	Environmental objectives→ Redevelopment performance	0.180	0.084	2.262	**	support
H3	Social needs→ Redevelopment performance	0.199	0.073	2.694	***	support
H4	Policy guidance→ Redevelopment performance	0.213	0.085	2.600	**	support
H5	Implementation conditions→ Redevelopment performance	0.171	0.072	2.297	**	support
H6	Policy guidance→ Economic incentives	0.283	0.087	3.607	0.015	support
H7	Policy guidance→ Implementation conditions	0.297	0.078	4.082	***	support
H8	Environmental objectives→ Implementation conditions	0.234	0.091	2.805	***	support
H10	Economic incentives→ Implementation conditions	0.205	0.092	2.301	***	support

Note: *** indicates p < 0.001; ** indicates p < 0.01.

), the pathways and effect strengths through which driving factors influence redevelopment performance have been further clarified. Notably, policy guidance demonstrates not only the most significant direct effect on redevelopment performance (β = 0.213) but also substantial indirect effects by stimulating economic incentives (β = 0.283) and enhancing implementation conditions (β = 0.297), indicating its central role within the overall driving system.

Both economic incentives (β = 0.196) and social needs (β = 0.199) exert significant positive effects on redevelopment performance. This suggests that for redevelopment to be successful, financial feasibility under market mechanisms and public expectations regarding quality of life and social equity must be simultaneously addressed, with both forming fundamental guarantees for project advancement and sustainable outcomes.

Further analysis reveals that economic incentives also indirectly promote redevelopment performance by improving implementation conditions (β = 0.205), reflecting their beneficial spillover effects on implementation conditions. Although environmental objectives (β = 0.180) and implementation conditions (β = 0.171) show relatively weaker direct effects, the former significantly influence the latter (β = 0.234), indicating that these factors primarily function as foundational elements that indirectly affect redevelopment performance by supporting other drivers and thus have an indispensable underlying role in the overall mechanism.

Given the use of ordered categorical data (5-point Likert scale), the model was re-estimated with the WLSMV estimator instead of the default ML to improve robustness. Analyses were performed in Mplus 8.3 using WLSMV, which provides more appropriate estimates for ordinal data. The results showed that under WLSMV estimation, the key path coefficient (economic incentives → redevelopment performance) shifted from 0.196 (p* < 0.001) to 0.205 (p* < 0.001), reflecting a marginal variation of 4.6%. All fit indices remained within satisfactory thresholds (χ²/df = 1.425, CFI = 0.952, RMSEA = 0.051). Following the application of the Satorra–Bentler scaled chi-square correction, the coefficient stabilized at 0.201 (p* < 0.001), confirming the consistency of results across different estimation methods.

4.6. FCM Model Development

SEM effectively identifies static relationships among variables in UILR, but it cannot capture dynamic processes or support scenario analysis. To address this limitation, this study introduces FCM as a complementary method. Specifically, path coefficients derived from SEM are used to quantify the causal connection weights among conceptual nodes in the FCM, establishing a statistically informed simulation network. This model enables dynamic process simulation and multi-scenario analysis of driving factors in UILR, thereby visualizing system dynamics and supporting policy simulation while maintaining causal objectivity.

The developed model simulates the dynamic impacts of driving factors on redevelopment performance. The conceptual nodes and their corresponding identifiers used in the model are summarized in

Table 7. Names and numbers of the nodes in the FCM mode.

Index	Variable Name	Index	Variable Name
C1	Economic incentives	C4	Policy guidance
C2	Environmental objectives	C5	Implementation conditions
C3	Social needs	CT	Redevelopment performance

. Specifically, nodes C1 through C5 represent causal factors (corresponding to the five driving factors), whereas CT serves as the target node (i.e., redevelopment effectiveness).

The initial weights were directly adopted from the path coefficients of the SEM shown in Table 6. Via Matrix W (Equation (6)), the causal relationship structure of the SEM was transformed and applied to FCM modeling, thereby constructing the corresponding FCM model.

$$\mathbf{W}\mathbf{=}\left[\begin{array}{cccccc}0& 0& 0& 0& 0.205& 0.196\\ 0& 0& 0& 0& 0.234& 0.180\\ 0& 0& 0& 0& 0& 0.199\\ 0.283& 0& 0& 0& 0.297& 0.213\\ 0& 0& 0& 0& 0& 0.171\\ 0& 0& 0& 0& 0& 0\end{array}\right]$$

(6)

5. Results of FCM Analysis

5.1. Predictive Analysis

In this study, predictive analysis is first conducted to systematically examine the independent effects of individual driving factors on redevelopment performance. The evolution of single driving factors is simulated under varying intervention levels while the dynamic response and final steady state of redevelopment performance (CT) are monitored.

Five simulation scenarios were designed to examine the effects of interventions on each driving factor Ci (i = 1, 2, 3, 4, 5). In each scenario, the driving factor Ci was evaluated using a 5-point linguistic scale with the following initial values: “−1” (highly unfavorable), “−0.5” (unfavorable), “0” (neutral), “0.5” (favorable), and “1” (highly favorable). The initial state of a specific driving factor was set to −1, −0.5, 0.5, or 1, representing the level of this driving factor after different degrees of intervention. The initial state of the rest of the driving factors was set to 0. Dynamic simulation yielded steady-state values for the target node CT. Figure 3 illustrates CT’s response to variations in each driving factor, with corresponding convergence values provided in

Table 8. Convergence values of outputs (CT) after several iterations in different scenarios.

	P (CT|Ci = 1)	P (CT|Ci = 0.5)	P (CT|Ci = −1)	P (CT|Ci = −0.5)
C1	0.9064	0.8164	−0.9064	−0.8164
C2	0.8302	0.7108	−0.8302	−0.7108
C3	0.7038	0.5854	−0.7038	−0.5854
C4	0.8803	0.8564	−0.8803	−0.8564
C5	0.7938	0.7055	−0.7938	−0.7055

.

The magnitude and nature of the influence exerted by each driving factor on CT can be assessed based on the trend of the CT curve and its steady-state convergence value. The results demonstrate a consistent positive correlation between all driving factors and CT: setting the initial values of the driving factors as positive (1 or 0.5) leads to positive CT convergence values, whereas setting the initial values as negative ( −1 or −0.5) results in negative convergence. For instance, when factor C1 is assigned initial values of 1 and 0.5, CT converges to 0.9064 and 0.8164, respectively; conversely, with initial values of −1 and −0.5, CT converges to −0.9064 and −0.8164, respectively. The complete dynamic process is depicted in Figure 3a, with the corresponding numerical results provided in Table 8.

Further analysis reveals that, when policy guidance (C4) is set at a “favorable” level (C4 = 0.5), its impact curve rises rapidly, approaching convergence by the fifth iteration with a value of 0.8564. This steep initial trajectory indicates that policy guidance can achieve rapid stabilization, demonstrating its nature as a short-term breakthrough driver. When C4 is set at a “highly favorable” level (C4 = 1.0), the convergence value reaches 0.8803, showing only marginal improvement compared to the “favorable” scenario. This suggests diminishing marginal returns beyond a certain threshold, implying that moderate policy intensity may suffice to achieve most objectives, thereby supporting cost-effective governance strategies.

In the case of economic incentives (C1), under “favorable” conditions (C1 = 0.5), the impact curve grows slightly slower than C4, converging around the tenth iteration at 0.8164. When the intervention intensity increases to C1 = 1.0, the convergence value rises to 0.9064. The larger incremental gain compared to C4 indicates substantial growth potential and returns to scale, revealing that economic incentives act as sustainable mid-term supports that benefit from continued investment.

Under an equivalent intervention intensity (C3 = 0.5), social needs (C3) exhibit the lowest convergence value (0.5854) and the slowest growth trajectory, reaching stability only around the fifteenth iteration. Despite its moderate ascent, the curve sustains an upward progression even as the influence of other driving factors diminishes. This pattern suggests that public engagement, community governance, and the development of social consensus represent protracted, cumulative processes, demonstrating the characteristics of a long-term dependent driver.

Under “favorable” interventions, the convergence values for environmental objectives (C2) and implementation condition (C5) are 0.8261 and 0.8155, respectively. Their influence intensities fall between those of C1 and C3, indicating their function as key supporting elements within the system.

5.2. Diagnostic Analysis

To further examine the sensitivity and contribution of individual driving factors to redevelopment performance (CT) variations, diagnostic analysis was conducted using FCM through reverse reasoning. This analysis addresses a key question: which driving factors require the most substantial adjustments when redevelopment performance (CT) targets are set at different levels?

The diagnostic results (Figure 4 and

Table 9. Convergence values of different influences (Ci) after several iterations in diagnostic analysis.

N	P (Ci|CT = 1.0)	P (Ci|CT = 0.5)	P (Ci|CT = −1.0)	P (Ci|CT = −0.5)
C1	0.9056	0.8734	−0.9056	−0.8374
C2	0.8370	0.7845	−0.8370	−0.7845
C3	0.7302	0.6108	−0.7302	−0.6108
C4	0.9280	0.9030	−0.9280	−0.9030
C5	0.8093	0.7495	−0.8093	−0.7495

) reveal a robust pattern: economic incentives (C1) and policy guidance (C4) serve as the two primary leverage points for regulating redevelopment performance. When CT targets the highest level (1.0), policy guidance (C4) achieves the maximum response value (0.9280); even at moderate target levels (CT = 0.5), its response value (0.9030) remains significantly higher than other factors. Notably, policy guidance (C4) demonstrates the steepest response curve and fastest convergence, confirming its effectiveness for rapid system transformation. Economic incentives (C1) exhibit the second-highest response (0.9056 at CT = 1.0), indicating that sustained high performance requires a strong economic underpinning. Conversely, social demands (C3) consistently yield the lowest responses (0.7302 at CT = 1.0), with the flattest curve and slowest convergence, representing a long-term cumulative force requiring persistent cultivation. Environmental objectives (C2) and implementation conditions (C5) exhibit intermediate response intensities and speeds, fulfilling auxiliary and synergistic roles.

In summary, diagnostic analysis confirms through reverse logic that to effectively enhance redevelopment performance, decision-makers should prioritize policy instruments while strengthening economic foundations and allowing for sufficient time for social effects to materialize. These findings align with the predictive analysis results, jointly providing empirical support for optimizing redevelopment strategies.

5.3. Hybrid Analysis

Hybrid analysis, the core component of this research, extends beyond static evaluation of single factors through constructing multiple intervention scenarios to systematically simulate the dynamic synergistic effects of different driving factor combinations on redevelopment performance (CT).

First, simulations of the evolutionary path of redevelopment performance under the most unfavorable conditions were conducted, where all driving factors and CT were set to minimum levels: C1 = −1, C2 = −1, C3 = −1, C4 = −1, C5 = −1, and CT = −1. After 20 iterations, the FCM model reached stability (Figure 5), with all driving factors remaining negative and redevelopment performance (CT) converging at −0.8936. These results indicate systemic failure, characterized by a negative feedback loop among driving factors, demonstrating both the necessity and urgency of external intervention.

Single-factor intervention strategies were designed to compare their effects on redevelopment performance (CT). In each strategy, only one factor’s initial state was elevated to 0.5 (“favorable”), while others remained at −1 (“highly unfavorable”). The simulation results (Figure 6) show that under these scenarios, all CT evolution curves ultimately converge to negative values, demonstrating that improving individual factors alone cannot fundamentally reverse the overall system state. Specifically, when only policy guidance (C4) was intervened with, the system responded most rapidly, achieving a significant improvement in CT in the short term. However, due to insufficient long-term momentum, it eventually converged to −0.8196. Intervention in economic incentives (C1) had a similar effect, but with a slightly lower convergence value. In addition, intervening solely in social demands (C3) yielded the worst improvement, achieving the lowest convergence value among all scenarios.

Multiple comprehensive intervention strategies were designed to compare how different combinations of driving factors affect redevelopment performance.

Table 10. State value settings for different integrated intervention scenarios and convergence values for redevelopment effectiveness.

Integrated Intervention Scenarios	C1	C2	C4	C5	CT	Effect Sorting
Integrated intervention scenarios 1	0.5	−1	0.5	−1	0.8590	4
Integrated intervention scenarios 2	−1	−1	0.5	0.5	0.8580	6
Integrated intervention scenarios 3	−1	0.5	0.5	−1	−0.8104	7
Integrated intervention scenarios 4	0.5	−1	−1	0.5	−0.8936	9
Integrated intervention scenarios 5	0.5	0.5	−1	−1	−0.8936	9
Integrated intervention scenarios 6	−1	0.5	−1	0.5	−0.8936	9
Integrated intervention scenarios 7	0.5	−1	0.5	0.5	0.8591	2
Integrated intervention scenarios 8	0.5	0.5	0.5	−1	0.8591	2
Integrated intervention scenarios 9	−1	0.5	0.5	0.5	0.8583	5
Integrated intervention scenarios 10	0.5	0.5	−1	0.5	−0.8934	8
Integrated intervention scenarios 11	0.5	0.5	0.5	0.5	0.8591	1

summarizes the state settings of driving factors (C1, C2, C4, C5) and corresponding CT convergence values across scenarios, with the dynamic trends shown in Figure 7.

According to Table 10 and Figure 7, all intervention scenarios that include both economic incentives (C1) and policy guidance (C4), specifically Scenarios 11, 8, 7, and 1, demonstrate exceptional effectiveness. Scenario 11, which represents the theoretical optimum through full-factor intervention, establishes the upper limit of redevelopment performance, with a convergence value of 0.8591. Scenarios 8 and 7, which are more practically significant, achieve convergence nearly identical to that of Scenario 11 while involving only three factors. This finding reveals that coordinating economic incentives (C1) with policy guidance (C4), supplemented by either environmental objectives (C2) or implementation conditions (C5), enables high-level redevelopment performance. The convergence value in Scenario 1 (0.8590), also closely approaching the upper limit, further confirms the decisive role of economic incentives (C1) and policy guidance (C4) within the system.

Furthermore, the hybrid analysis uncovers a determining principle: policy guidance (C4) functions as the critical “switch” for activating the redevelopment system. As evidenced by Scenarios 4, 5, 6, and 10, all intervention combinations that exclude policy guidance (C4) yield very negative CT convergence values, even when simultaneously engaging economic incentives (C1), environmental objectives (C2), and implementation conditions (C5). This demonstrates that in the absence of policy direction and institutional approval, other driving factors are not sufficient to overcome the system’s inherent development inertia.

Additionally, in Scenarios 2 (C4 and C5 intervention) and 9 (C4, C5, and C2 intervention), CT converged at values of 0.8580 and 0.8583, respectively, indicating a strong positive performance. These outcomes demonstrate that policy guidance (C4) creates effective synergy with implementation conditions (C5), maintaining a satisfactory redevelopment performance even when economic incentives (C1) remain unfavorable. By contrast, Scenario 3 (intervention in C4 and C2) resulted in CT converging −0.8104, reflecting a persistently negative system performance. This suggests that, in the absence of implementation conditions (C5), environmental objectives (C2) primarily increase the costs and limit the development intensity, failing to synergize with policy guidance (C4) and leading to systemic inertia. Together, these scenario comparisons highlight that while policy guidance is essential for system activation, its effectiveness fundamentally relies on support from the implementation conditions.

6. Discussion

6.1. Static Structural Relationships

Static analysis based on SEM confirms the theoretical validity of the multidimensional driving framework. The results indicate that policy guidance functions as the core hub of the entire driving system: it produces the strongest direct effect on redevelopment performance and generates indirect effects through either economic incentives or implementation conditions. This finding clarifies the structural role of policy guidance as the master switch; it directly sets rules and direction while simultaneously activating market forces and strengthening implementation conditions to reshape value realization pathways. In addition, both environmental objectives and economic incentives influence redevelopment performance indirectly through the shared mediator of implementation conditions. Environmental objectives drive the improvement of implementation conditions by raising technical standards and tightening ecological oversight, whereas economic incentives provide critical support for land consolidation and infrastructure provision through capital investment and market momentum. Thus, the two drivers enhance redevelopment performance indirectly via green regulation and capital empowerment, respectively.

In contrast, economic incentives exert no statistically significant direct effect on environmental objectives (Hypothesis 9 not supported), suggesting their relationship is not a simple linear positive correlation but rather manifests as a dynamic coupling. The specific expression of this relationship depends on the developmental stage and the institutional context: during extensive development phases, the two typically represent a trade-off, as economic expansion intensifies environmental pressures; during transitional phases, however, technological advancement and institutional regulation can foster synergistic development, creating win–win scenarios where economic returns support environmental improvements.

6.2. Dynamic Evolutionary Relationships

Dynamic simulation based on FCM further reveals the temporal characteristics of a driving factor, adding a temporal dimension to the static structural relationships. The findings demonstrate that no single-factor intervention, including interventions via the most influential factor—policy guidance—is sufficient to release the system from being trapped in a low-efficiency state. As a complex social–economic–environmental system, UILR is characterized by tight coupling and recursive feedback, meaning that isolated improvements in any one element cannot disrupt the existing equilibrium. Only when multiple key factors act in concert can sufficient momentum be generated to shift the system toward a high-efficiency development regime. Among these factors, policy guidance plays an irreplaceable switching role: in its absence, the system remains trapped in a negative state even when other critical elements are present. This result underscores the unique status of policy guidance as a necessary condition for activating the system, reflecting the characteristics of China’s urban governance system while corroborating the premise of institutional theory: “rules define behavioral boundaries” [61]. Notably, policy guidance combined with implementation conditions creates an effective “decision–execution” chain that mitigates economic incentive deficiencies, offering viable pathways for UILR during economic transitions or when there are resource constraints.

6.3. Limitations of the Study

Several limitations should be noted. First, this study focuses primarily on five core driving factors; however, in reality, potential variables such as cultural characteristics within regions, micro-level behavioral decisions, and the impact of sudden events may also significantly influence the system. Future research could incorporate more diverse factors to construct a more comprehensive driving framework. Second, although the integrated SEM and FCM approach is effective in examining multivariate relationships and system dynamics, it is not without its methodological constraints. SEM relies on strong statistical assumptions (e.g., linearity and normality), which often diverge from the nonlinear and path-dependent nature of real-world systems. Furthermore, causal inferences drawn from SEM remain fundamentally statistical associations and are susceptible to endogeneity bias due to omitted variables (e.g., local institutional differences). In FCM, network weights depend on SEM outputs, and its linear superposition rule may inadequately capture emergent properties and nonequilibrium system evolution. Subsequent studies could explore the introduction of Nonlinear Structural Equation Modeling (NSEM) [62] or System Dynamics (SD) [39] methods to more accurately simulate feedback mechanisms, delay effects, and critical transition behaviors within the system. Finally, the SEM measurement system and pathway data were derived from expert surveys in eight cities within China’s Yangtze River Delta, grounding the analysis in China’s specific institutional and land-governance context. Caution is therefore warranted when extending the findings to other regions, particularly those with differing land-rights structures, policy implementation, and market actor roles. Future work could employ cross-regional and cross-type comparative studies to further validate the context dependency and evolutionary patterns of the driving mechanisms.

7. Conclusions

Building on Yang’s triple-structure theory of development dynamics, this study incorporates multiple driving factors within a unified “drivers–pressures–enablers” analytical framework and integrates SEM with FCM to develop a comprehensive methodology that combines static structural analysis with dynamic scenario simulation.

The findings reveal a synergistic evolution among economic incentives, environmental goals, social needs, policy guidance, and implementation conditions. This process unfolds sequentially as “policy-driven breakthroughs, sustained economic impetus, and long-term societal cultivation,” highlighting the path-dependent nature of UILR in China. The developed SEM–FCM hybrid model effectively integrates static validation with dynamic simulation, overcoming the limitations of conventional approaches in capturing the nonlinearity, feedback effects and temporal heterogeneity in complex systems, offering robust methodological support for understanding the operational principles of UILR as a complex system. Furthermore, the model can be extended into an integrated “scenario simulation–policy evaluation” platform to assist decision-makers in simulating the long-term effects of different strategy combinations, optimizing the selection and timing of policy instruments, and advancing high-quality urban stock renewal alongside governance modernization.