Spatiotemporal Evolution and Driving Factors of Tourism Eco-Efficiency: A Three-Stage Super-Efficiency SBM Approach

Abstract

Tourism ecological efficiency (TEE) is a significant indicator of the development level of green and intensive tourism. However, conventional directional and radial TEE measurement approaches overlook critical factors such as intermediate process influences and input–output slack variables, potentially leading to biased estimates. Urban areas are key to coordinating tourism across provinces, so accurately assessing the TEE is vital for sustainable regional tourism. This study uses an improved TEE measurement model to measure the TEE of the Guangdong–Fujian–Zhejiang (GFZ) coastal city clusters from 2010 to 2021. The improved TEE measurement model is a three-stage super-efficiency SBM approach. It then uses standard deviation ellipses and geographic detectors to analyze the TEE’s spatiotemporal characteristics and influencing factors. The findings indicate the following: (1) The three-stage super-efficiency SBM approach improves the accuracy and validity of measurement results by removing external environmental variables. (2) During the study period, the TEE values of the GFZ coastal city clusters were above average (except for Meizhou, where the efficiency improved). Temporally, the TEE values of 75% of the cities showed an increasing trend; spatially, the high-value areas increased significantly, the middle- and low-value areas decreased, and the center of gravity shifted to the north and south. (3) The years 2016–2021 saw an increase in external development factors and the use of external resources. The study’s findings can serve as scientific benchmarks for TEE measurement, as well as the low-carbon and environmentally friendly growth of tourism in urban agglomerations.

1. Introduction

TEE is the degree to which tourism and ecosystems are developed in harmony, with the goal of maximizing benefits while minimizing damage [1,2]. Enhancing the TEE helps the tourism industry develop sustainably and promote harmonious coexistence between people and nature. It is also essential for the tourism industry’s green and low-carbon transition [3]. Huppes and Ishikawa (2005) introduced the idea of ecological efficiency (EE), which is concerned with assessing the environmental impact of economic products [4]. Gössling and colleagues introduced EE into the tourist field in 2005, which, in turn, encouraged many researchers to conduct studies on TEE [5,6]. The development of TEE measurement techniques [7], investigation of spatiotemporal features [8], and identification of significant factors [9] have all been included in this research. The predominant methodologies employed for TEE calculations encompass the single-indicator method [5,10], carbon accounting with ecological carrying capacity metrics as a standardized quantification framework [7], and the data envelopment analysis (DEA) method [11]. Traditional DEA models ignore intermediate links’ impacts on efficiency and the slack between inputs and outputs. Therefore, non-angle and non-radial super-efficient SBM approaches with non-expected outputs have been proposed [12]. Nevertheless, the SBM-derived efficiency scores lack inherent guidance for performance enhancement, mandating auxiliary techniques, such as regression modeling or qualitative diagnostics. To exclude the impact of external variables and random errors on TEE measurements, the three-stage DEA model [13] was proposed. This model integrates the DEA and stochastic frontier gravity models. However, integrating non-desired outputs and external environmental variables into three-stage super-efficiency SBM approach remains underutilized, consequently compromising the TEE measurement accuracy.

TEE is subject to multiple threats from political, economic, and environmental factors, and existing studies have analyzed the impact of TEE due to internal drivers [14,15]. As for external drivers, some studies have found that environmental governance and the number of patents have a positive effect on TEE, while the structure of the tourism industry and fiscal expenditure have a negative effect on TEE [16,17]. Concerning the efficacy of regional tourism development and its constituent dimensions—tourism enterprise management practices, itinerary design, product development, and visitor experience—exogenous drivers exhibit broader influence across these facets compared with their endogenous counterparts. This inclusive advantage is especially significant in the overall impact mechanism of TEE [18]. Conventional econometric approaches applied in prior TEE determinant studies risk estimation bias from multicollinearity data while inadequately modeling driver interdependencies [19]. Though scholars have begun to introduce geographic detector models into the identification of influencing factors of TEE [19,20], most of them were studied from the provincial or large-scale region perspective [21,22], and fewer have explored the factors influencing TEE in urban agglomerations from the perspective of prefectural-level cities.

As a strategic functional area for China’s inter-provincial economic cooperation and international integration, the coastal city cluster of Guangdong, Fujian, and Zhejiang has gradually become an integral component of China’s coastal travel destination belt, and its tourism industry now serves as a key contributor to local economic expansion. However, while the rapid development of tourism is underway, the problems of ecological and resource environment deterioration and spatial imbalance in utilization caused by tourism development and activities have resulted in coastal city tourism presenting high resource demand. However, surging tourism expansion, coupled with the resultant degradation of ecosystems and natural resources from tourism-related pressures, has led to the imbalance of high resource consumption and low output, which has become a real dilemma for sustainable tourism advancement in the GFZ coastal city clusters [11]. Existing research on regional TEE tends to focus on a single spatial scale, such as a province or a single tourist city, and lacks analysis of the coordination and competition within urban agglomerations that span administrative boundaries. Coastal urban agglomerations are characterized by both ecological fragility and economic extroversion [23,24]. Compared with inland cities, coastal urban agglomerations face more severe resource and environmental constraints, meaning the coordination of regional sustainable development is of urgent theoretical and practical significance [25].

Concisely stated, attention in the current research is directed toward the following key aims: (1) accurately measuring the TEE for the urban agglomerations in Southeastern Coastal China based on the three-stage super-efficient SBM approach; (2) exploring the temporal and spatial evolutions of the TEE of selected coastal city clusters in Southeast China (GFZ) based on the TEE measurement results; and (3) applying a geographical detector to analyze major contributors to the TEE. The present study’s objective is to provide a scientific basis for improving the TEE of urban agglomerations and contributing to regional sustainability.

2. Data Sources and Methods

2.1. Indicator System Construction

TEE is regarded as a key indicator of tourism resource allocation efficiency. It aims to mitigate the negative impacts on the environment and promote the development of society and ecology. Based on this and drawing on existing studies [11,22,26], the measurement indicators of TEE are divided into two main categories (inputs and outputs), in which the input indicators are selected as the quantity of high-tier tourist attractions, the count of classified hotels, and physical assets within the tourism sector, alongside the labor force. With reference to the related research, the value is derived through the multiplication of the fixed capital stock within service industries. The industry is characterized by its reliance on tourism as a primary economic driver (share of tourism = total income from tourism/GNP) [27], and the labor force inputs are chosen to characterize the tourism workforce size, with the same calculation method as above [28]. The output indicators are the total tourism revenue and total tourism receipts, which are two variables that visually reflect the economic benefits of tourism as quantitative indicators to measure the desired output [29]. Carbon emissions from tourism are often used as an indicator of undesired output [30,31,32]. However, tourism wastewater emissions, tourism sulfur dioxide emissions, and tourism soot and dust emissions were selected to characterize the non-desired outputs, considering data availability. Drawing on [33], the tourism share ratio was used for conversion, tourism wastewater emissions = wastewater emissions × tourism share ratio, and tourism-related SO₂ discharges and particulate matter outputs were computed employing identical methodologies.

Thus, the TEE evaluation index system was constructed (

Table 1. Evaluation index system of TEE.

Type	Indicator System	Specific Indicators	Unit of Measure
Input indicators	Resources	Number of A-class scenic spots	Classifier for individuals or groups of individuals
	Services	Number of star-rated hotels	Units
	Capital	Fixed capital stock of tourism	Billion
	Labor	Employees in the tourism industry	Million
Output indicators	Desired output	Gross tourism income	Billion CNY
		Total tourism reception	Million
	Non-expected outputs	Tourism wastewater emissions	Million tons
		Sulfur dioxide emissions from tourism	Tons
		Tourism fume and dust emissions	Tons

).

External environmental variables, i.e., factors that have an impact on TEE and cannot be subjectively changed or grasped within a short period [34,35], were mainly selected in existing studies as follows: politics, economy, industrial structure, science and technology, and environment. This study makes use of external environment factors [36], including government spending as a percentage of GNP (intervention); industrial structure’s proportion of the GDP in the GNP; the GNP per capita (economic development level) [37]; and urban greening, as measured by the green urban area [38].

2.2. Data Sources

This research primarily utilized national socioeconomic reports and municipal yearbooks from multiple cities across Guangdong, Fujian, and Zhejiang Provinces. Additionally, the official statistical data from municipal statistics offices at the provincial and prefectural levels was checked. Provincial and prefectural tourism bureau websites provided A-level scenic area listings and star-rated hotel counts. Regarding the data completeness, the study assessed the coverage of the acquired dataset and confirmed that it basically covered the period and geographic units required for the study (all prefectural-level cities in the city cluster). For the small amount of missing data in individual prefectures and cities in a given year (mainly related to the data of A-grade scenic spots and star-rated hotels), which accounted for about 2.5% of the total sample size, the linear interpolation method was used. The method assumes continuity of time-series data, where the missing data are estimated using valid data from adjacent years before and after the missing values. All data sources and data-processing steps (including missing value identification and interpolation) are documented in detail to ensure the study’s reproducibility [39].

2.3. Processing and Methods

2.3.1. Three-Stage Super-Efficiency SBM Approach

Currently, much academic research has evaluated EE with the help of traditional DEA and the super-efficiency approach; however, the efficacy of these methodologies in eradicating the influences of random errors and extraneous environmental factors on EE is questionable. Moreover, their inability to accurately depict the true state of EE is a salient concern. This study aims to overcome previous studies’ shortcomings and use the three-stage super-efficiency SBM approach to obtain a more objective and realistic level of TEE. This will provide a scientific basis for TEE improvement policies in the GFZ coastal city clusters [40,41].

• Stage I: Super-Efficient SBM Approach

There are complex input and output elements in the TEE production process, and the super-efficient SBM approach can effectively solve the problem of multiple inputs and multiple outputs while considering undesired outputs and input–output slackness. The SBM approach is used to measure the TEE of the GFZ coastal city clusters from 2010 to 2021. Its hierarchical classification is based on the study by Charnes et al. [42] and is divided into low-value zones: TEE ≤ 0.6, medium-value zones: 0.6 < TEE ≤ 0.8, and high-value zones: TEE > 0.8.

The formula appears below,

$$\mathit{min}\rho ={\displaystyle \frac{\frac{1}{m}\sum _{i=1}^m\frac{\overline{x}}{x_{ik}}}{\frac{1}{r_1+r_2}\left(\sum _{s=1}^{r_1}\frac{y^{\overline{d}}}{y_{sk}^d}+\sum _{q=1}^{r_2}\frac{y^{\overline{u}}}{y_{qk}^u}\right)}}$$

(1)

$$\left\{\begin{array}{c}\overline{x}\ge {\displaystyle \sum _{j=1,\ne k}^n}{x}_{ij}\lambda ; y^{\overline{d}}\le {\displaystyle \sum _{j=1,\ne k}^n}{y}_{sj}^d{\lambda }_j\\ \overline{y^d}\ge {\displaystyle \sum _{j=1,\ne k}^n}{y}_{qj}^d{\lambda }_j; \overline{x}\ge x_k; \overline{y^d}\le y_k^d; \overline{y^u}\ge y_k^u; {\lambda }_{ij}\ge 0\\ i=1,2,\dots ,m; j=1,2,\dots n; s=1,2,\dots ,r_1; q=1,2,\dots ,r_2\end{array}\right.$$

where $\mathsf{\rho }$ represents the TEE; $\mathrm{n}$ represents the decision unit; $\mathrm{m}$ represents the inputs for the municipality TEE; ${\mathrm{r}}_1$ is the desired output; ${\mathrm{r}}_2$ is the undesired output; $\mathrm{x}$, ${\mathrm{y}}^{\mathrm{d}}$, and ${\mathrm{y}}^{\mathrm{u}}$ present the elements of the corresponding matrix; and ${\mathrm{\lambda }}_{\mathrm{ij}}$ represents the weights assigned to the decision unit.

2.

Stage II: Stochastic Frontier Gravity Model SFA

Effectively identifying and separating redundant decision-making units caused by managerial inefficiency involves constructing a multiple regression model similar to the SFA with input slack variables as explanatory variables and external environment variables as explanatory variables; the specific modeling and formulation of this stage is detailed in [40,43].

3.

Stage III: Super-Efficient SBM Approach

The adjusted input indicator data of the DMU is integrated with the original output indicator data and re-measured with the super-efficient SBM approach. Currently, input indicator data are stripped of external environmental variables and random errors, making them more objective and accurate [44]. The flowchart of the three-stage super-efficiency model is shown in Figure 1.

2.3.2. Standard Deviation Ellipse

The standard ellipse deviation (SED) model is a statistical analysis method designed to characterize the distribution of data. The method summarizes the distribution of data points by constructing a standardized ellipse, whose long and short axes, as well as its flatness, reveal the center of agglomeration and the data’s dispersion level, respectively. This method is suitable for analyzing the characteristics of geographic detector elements that vary in time and space. Its mathematical formulae are detailed in [45].

2.3.3. Geographical Detector

A geo-detector is a statistical analytical instrument whose primary function is to examine the factors that influence the distribution of geographic detector models. There are four core components embedded in this modeling system, namely, the ecological, risk, factor, and interaction detectors [20]. The interaction detector is employed by comparing the extent to which the dependent variable is explained using multiple drivers, as opposed to the extent to which it is explained by a single factor. There are five types of two-factor X1 and X2 interactions: nonlinearly attenuated, two-factor enhanced, one-factor nonlinearly attenuated, nonlinearly enhanced, and mutually independent [19]. Therefore, this study employed a factor detector to assess the explanatory properties of the TEE factor. Using an interaction detector, we aimed to analyze in depth the interactions between factors influencing the TEE of each prefecture-level city within the coastal urban clusters of Guangdong, Fujian, and Zhejiang, quantifying their influence degree with the following formula:

$$q_D=1-{\displaystyle \frac{\sum _{i=1}^m{N}_i{\sigma }_i^2}{N{\sigma }^2}}$$

(2)

where N and ${\mathrm{N}}_{\mathrm{i}}$ denote the number of study units and sub-study units, respectively; m denotes the number of influencing factors; and ${\mathsf{\sigma }}^2 \mathrm{and} {\mathsf{\sigma }}_{\mathrm{i}}^2$ denote the variance of the efficiency values in the study area and sub-study area, respectively. ${\mathrm{q}}_{\mathrm{D}}\in \left\{0,1\right\},$ where a larger ${\mathrm{q}}_{\mathrm{D}}$ indicates that the influencing factors have a stronger explanatory power for the TEE changes, and vice versa; a value of 1 for ${\mathrm{q}}_{\mathrm{D}}$ indicates that the influencing factor D completely controls the changes in the TEE pattern, and a ${\mathrm{q}}_{\mathrm{D}}$ value of 0 indicates that the influencing factor is completely irrelevant to the TEE changes. In selecting the TEE influencing factors, the study referred to [46] and, combined with the accessibility of the data, finally established six indicators: economic development, industrial structure, transportation, urban green space area, government intervention, and urbanization level (

Table 2. Influencing factors of TEE.

Level 1 Indicators	Level 2 Indicators	Specific Indicators	Unit
Economic development	Tourism economics	GDP per capital	CNY
Industrial structure	Structure of the tourism industry	Proportion of total tourism Revenue to tertiary industry output value	Percent
External environment	Traffic	Road passenger traffic volume	Million
	Government intervention	Fiscal expenditure	Percent
	Urbanization level	Proportion of urban population	Percent
	Greening level	Urban green space area	Hectares

).

3. Results

3.1. Results of TEE Measurement in the GFZ Coastal City Clusters Based on the Three-Stage Super-Efficient SBM Approach

3.1.1. Stage I

To evaluate TEE across prefecture-level cities within the GFZ coastal city clusters during 2010–2021, the study employ a three-stage super-efficiency framework. The initial stage employs a super-efficiency slack-based measure (SBM) approach, with corresponding outcomes documented in

Table 3. Value of TEE in the first stage of the GFZ coastal city clusters from 2010 to 2021.

City	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021
Fuzhou	0.088	0.095	0.099	0.107	0.128	0.141	0.189	0.240	0.306	1.001	0.233	0.207
Xiamen	0.150	0.127	0.151	0.200	0.187	0.250	0.440	0.583	0.830	1.035	1.002	1.060
Putian	0.058	0.059	0.056	0.087	0.107	0.119	0.150	0.355	0.601	1.053	1.007	1.051
Sanming	0.045	0.062	0.067	0.079	0.101	0.115	0.127	0.219	0.299	0.564	1.030	1.015
Quanzhou	0.137	0.141	0.159	0.165	0.178	0.182	0.202	0.266	0.321	0.361	0.200	0.196
Zhangzhou	0.071	0.080	0.084	0.079	0.093	0.103	0.119	0.141	0.171	0.200	0.142	0.252
Nanping	0.086	0.154	0.174	0.179	0.207	0.239	0.353	0.557	0.827	1.051	1.034	1.019
Longyan	0.073	0.082	0.100	0.115	0.153	0.162	0.201	0.240	0.381	0.436	0.405	0.518
Ningde	0.053	0.085	0.092	0.139	0.169	0.205	0.265	0.271	0.468	1.001	0.671	1.205
Wenzhou	0.129	0.118	0.137	0.159	0.189	0.198	0.330	0.502	1.077	1.011	1.011	1.06
Lishui	0.195	0.231	0.277	0.288	0.323	0.388	0.481	0.597	1.001	1.047	1.118	1.136
Quzhou	0.080	0.104	0.113	0.139	0.181	0.244	0.322	0.633	1.010	1.018	0.782	1.067
Shantou	0.088	0.093	0.105	0.111	0.134	0.204	0.408	0.698	1.010	1.057	0.276	0.560
Chaozhou	0.109	0.120	0.134	0.144	0.217	0.245	0.308	0.398	0.602	1.089	0.520	0.218
Jieyang	0.072	0.099	0.111	0.115	0.135	0.151	0.194	0.287	1.061	1.111	0.733	1.184
Meizhou	0.148	0.262	0.248	1.028	0.465	1.021	1.004	0.442	0.457	0.527	0.335	1.076

.

Without considering the impact of external environmental variables and random errors, the overall trend of TEE in each prefecture-level city from 2010 to 2021 shows a fluctuating upward trend. However, the overall TEE levels of each prefecture-level city remain relatively low, at or below the medium level. This suggests that prefecture-level cities within the GFZ coastal city clusters may generally face issues such as unreasonable allocation of production inputs and outputs, an unbalanced industrial structure, low resource utilization rates, and insufficient environmental protection efforts.

3.1.2. Stage II

The input slack variable derived from the first-stage super-efficient SBM approach constituted the dependent variable, with four exogenous determinants acting as independent variables in the subsequent stochastic frontier gravity model analysis. Results are summarized in

Table 4. Results of regression analysis of the second stage stochastic frontier gravity model.

Item	Resource Input Slack Variables	Service Input Slack Variables	Capital Input Slack Variables	Labor Input Slack Variables
Constant term	–8.148 ***	22.43 ***	–22.60 ***	–14.54 ***
Economic development level	1.197 ***	–0.470 ***	1.344 ***	0.339 ***
Industrial structure	1.084 ***	–0.321 **	3.345 ***	2.904 ***
Government intervention	0.232 **	0.108 **	0.391 ***	0.422 ***
Urban greening Area	–0.694 ***	1.113	0.0433	0.165 ***
Sigma2	0.251	1.564	1.596	11.700
Gamma	0.646	0.987	0.884	0.986
Unilateral LR test value	7.31 ***	0.74	8.29 ***	6.96 ***

*** p < 0.01 and ** p < 0.05.

.

The slack variable test indicates that there is no managerial inefficiency in service inputs (0.74, not significant), while there is significant managerial inefficiency in resource, capital, and labor inputs, necessitating the removal of external environmental variables. A Gamma value > 0.6 confirms the rationality of variable selection. The second-stage analysis indicates that economic development significantly increases resource, capital, and labor redundancy (at the 1% level) but reduces service redundancy; the rise in the tertiary industry’s share exacerbates resource element redundancy (at the 1% level) but reduces service redundancy (at the 5% level), reflecting low-efficiency issues in the utilization of elements within the tertiary industry. Government intervention has a significant positive impact on all input redundancies (at the 1%/5% level), revealing the need to shift toward market-based regulatory mechanisms to address inefficiency. Urban greening, while improving resource utilization efficiency (significantly reducing surplus at the 1% level), significantly increases labor demand and service capital investment, leading to new inefficiency issues.

3.1.3. Stage III

Using a highly efficient measurement method based on relaxation, the third stage recalculated the TEEs of each prefecture-level city in the GFZ coastal city clusters from 2010 to 2021. By using adjusted inputs and unmodified outputs, adjusted TEE values were obtained (

Table 5. Tourism eco-efficiency value of the third stage of GFZ coastal city clusters from 2010 to 2021.

City	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021
Fuzhou	0.499	0.496	0.501	0.508	0.515	0.525	0.548	0.572	0.610	0.639	0.644	0.782
Xiamen	1.007	0.866	0.791	0.758	0.726	0.719	0.739	0.790	1.002	0.888	1.003	1.100
Putian	1.078	1.009	1.001	1.010	1.001	0.958	0.953	1.012	1.005	1.019	0.950	1.034
Sanming	0.719	0.732	0.727	0.721	0.698	0.679	0.676	0.675	0.679	0.683	0.623	0.700
Quanzhou	0.390	0.392	0.404	0.415	0.427	0.441	0.460	0.487	0.505	0.523	0.498	0.531
Zhangzhou	0.743	0.691	0.698	0.661	0.639	0.630	0.641	0.648	0.646	0.664	0.621	0.745
Nanping	1.007	1.002	0.963	0.880	0.819	0.777	0.784	0.784	0.793	0.793	0.727	0.717
Longyan	0.767	0.738	0.739	0.715	0.700	0.672	0.665	0.667	0.678	0.689	0.666	0.803
Ningde	1.005	1.003	0.966	1.005	0.980	0.932	1.001	0.882	1.004	1.002	0.857	1.049
Wenzhou	0.406	0.411	0.424	0.437	0.457	0.474	0.501	0.520	0.550	0.547	0.570	0.549
Lishui	0.666	0.621	0.598	0.576	0.563	0.560	0.577	0.584	0.585	0.603	0.616	0.586
Quzhou	0.751	0.729	0.718	0.696	0.667	0.640	0.631	0.653	0.651	0.684	0.735	1.019
Shantou	0.443	0.452	0.456	0.459	0.470	0.496	0.519	0.542	0.563	0.629	0.608	0.616
Chaozhou	0.786	0.800	0.804	0.792	0.788	0.780	0.782	0.790	0.836	1.008	1.016	1.109
Jieyang	0.479	0.498	0.508	0.511	0.524	0.548	0.583	0.748	1.050	1.034	0.509	1.460
Meizhou	0.375	0.405	0.419	0.430	0.447	0.462	0.479	0.498	0.523	0.570	0.442	0.517
Average Value	0.695	0.678	0.670	0.661	0.651	0.643	0.659	0.678	0.730	0.748	0.693	0.782

).

After the third stage of adjustments, the TEE of each prefecture-level city in the GFZ coastal city clusters from 2010 to 2021 increased significantly compared to before the adjustments. After eliminating the influence of external environmental variables and random errors, the TEE distribution of each region was uniform and more reasonable than before the adjustments.

Looking at the prefecture-level cities in the GFZ coastal city clusters, the TEE indicators of 15 prefecture-level cities have significantly improved. In the first stage, only a few cities like Xiamen and Putian achieved ultra-high efficiency (TEE > 1.0). By the third stage, a stable high-efficiency cluster had formed, with five cities—Putian, Xiamen, Ningde, etc.—consistently maintaining high values (>0.8). Notably, Putian’s TEE surpassed 1.0 after nine years, reflecting the true efficiency emerging after environmental variables were stripped away. The “pseudo-high efficiency” phenomenon prevalent in the first stage, such as Meizhou’s sudden increase to 1.028 in 2013, disappeared in the third stage after adjustment, with its 2013 value dropping to 0.430 after correction, confirming the severe distortion of original efficiency by external environmental factors; under the impact of the pandemic, many regions experienced abnormal efficiency surges in the first stage (e.g., Sanming reached 1.030 in 2020), but all saw consistent declines in the third stage (Sanming dropped to 0.623), highlighting the correction model’s ability to filter out random noise; Chaozhou City experienced significant fluctuations in efficiency in the first stage (plummeting to 0.218 in 2021), but it continued to rise in the third stage to 1.109, proving that its early inefficiency was actually due to environmental disadvantages, and its true ecological tourism synergy efficiency was severely underestimated. The model correction eliminated the “efficiency bubble” and provided a reliable benchmark for policy-making.

3.2. Characteristics of Time-Series Changes of TEE in the GFZ Coastal City Clusters

A line chart was created using the TEE values for the coastal cities of Guangdong, Fujian, and Zhejiang from 2010 to 2021 to visually illustrate the changes, as shown in Figure 2. The trends in TEE for each city in the coastal cities of Guangdong, Fujian, and Zhejiang from 2010 to 2021 varied. In terms of the average TEE values per year, the overall trend shows a fluctuating upward trend (Figure 2d).

As depicted in Figure 2a–c, Jieyang exhibited the most pronounced variability, followed by Nanping and Xiamen, whereas other urban centers demonstrated comparatively stable trajectories during 2010–2021. Nine cities—Putian, Sanming, Nanping, Lishui, Xiamen, Longyan, Quzhou, Zhangzhou, and Ningde—initially displayed declining trends. Subsequently, Xiamen, Longyan, Quzhou, Zhangzhou, and Ningde transitioned to fluctuating upward trajectories, yielding an overall positive tendency. Six urban centers (Ganzhou, Quanzhou, Wenzhou, Shantou, Chaozhou, and Meizhou) manifested sustained upward momentum with oscillatory characteristics. The efficiency of the three provinces shows differentiated paths for Fujian (resource-based), Zhejiang (technology-based), and Guangdong (outward-oriented), all pointing to the evolutionary pattern of “from internal accumulation to external coordination.” Collectively, 75% of the cities studied achieved net TEE growth, signifying enhanced efficacy in ecological conservation measures.

3.3. Characteristics of Spatial Evolution of TEE in the GFZ Coastal City Clusters

The TEE values of each city in the GFZ coastal city clusters from 2010 to 2021 were divided into three levels according to the efficiency level: high-value area (≥0.8), higher-value area (0.6~< 0.8), and medium-low-value area (0~< 0.6), and were visualized with the help of ArcGIS 10.8 software to analyze the regional differences and spatial evolution characteristics. Based on China’s key policy and cycle planning nodes, five years were selected to map the TEE distribution: 2010 (the end of the 11th Five-Year Plan), 2013 (the middle of the 12th Five-Year Plan), 2016 (during the 13th Five-Year Plan), 2019 (the end of the 13th Five-Year Plan), and 2021 (during the 14th Five-Year Plan).

Figure 3 shows that in the coastal urban agglomeration of the GFZ coastal city clusters in 2010, Wenzhou, Fuzhou, Quanzhou, Shantou, Jieyang, and Meizhou were in the low-medium value zone; Quzhou, Lishui, Sanming, Longyan, Zhangzhou, and Chaozhou were in the higher-value zone; and Nanping, Ningde, Putian, and Xiamen were in the high-value zone. In 2013, Lishui City receded from the higher-value zone to the low-middle-value zone, Xiamen City receded from the high-value zone to the higher-value zone, and the other cities remained unchanged; Nanping, Ningde, Putian, and Xiamen constituted the premium tier. By 2013, Lishui transitioned downward from the upper stratum to the baseline category, while Xiamen shifted from the premium tier to the upper stratum, with the other urban centers maintaining their prior classifications. Subsequent transitions included Nanping’s demotion to the upper stratum (2016); Lishui’s recovery to the upper stratum, Fuzhou’s elevation from the baseline to upper stratum, Xiamen’s reinstatement to the premium tier, Chaozhou and Jieyang’s ascension to the premium tier in 2019, and Shantou’s advancement to the upper stratum. In 2021, the TEE indicators of Longyan City and Quzhou City rose to the high-level zone, while Lishui City fell back to the medium-low-value zone. Data analysis revealed that Lishui City’s tourism-related SO2 emissions rebounded from previous years, while the desired output indicator showed a downward trend. It indicates that its actual efficiency was more constrained by internal management or technical factors rather than external environmental disadvantages. Under these compounded influences, Lishui’s TEE demonstrated an initial ascent followed by descent.

During 2010–2016, the study area exhibited a latitudinal efficiency gradient decreasing coastward, whereas the 2016–2021 period revealed an inverted spatial progression with generally coast-inclining values. Comparative analysis of the 2010 versus 2021 TEE spatial distributions indicates marked expansion of high-performance zones (>0.8), alongside contraction of the intermediate- and low-value regions (<0.6), evidencing progressive enhancement of the regional TEE.

To deeply explore the spatial movement patterns and evolutionary paths of TEE in the GFZ coastal city clusters, this study used ArcGIS software to carry out the standard deviation ellipse and center of gravity movement analyses. Figure 4 shows the following: (1) During the study period, the TEE standard deviation ellipse of the GFZ coastal city clusters mainly exhibited a spatial distribution pattern in the direction of “northeast–southwest”. (2) The area covered by the standard deviation ellipse of TEE in the GFZ coastal city clusters showed a trend of gradual growth over time, although the number of cities involved remained constant. (3) In this coastal urban agglomeration, the long half-axis of the TEE standard deviation ellipse was always longer than the short half-axis, and the short half-axis had a relatively small change, while the long half-axis showed an expanding tendency, which indicates that the TEE gradually spread out in the northeast–southwest direction, while it remained relatively stable in the northwest–southeast direction. (4) The migration of the TEE center of gravity in the GFZ coastal city clusters mainly occurred in the region bordering Sanming City and Quanzhou City, showing a significant northeast–southwest shifting trend. Over time, the influence of the eastern coastal cities in terms of tourism management experience and low-carbon environmental awareness gradually spread to the central and western regions, and the driving effect of the TEE became increasingly prominent [47]. In particular, the TEE’s center of gravity moved southwestward from Quanzhou City between 2010 and 2013; it continued to move southwestward from 2013 to 2016, though the shift was not very significant. The center of gravity changed and moved to Sanming City’s jurisdiction between 2016 and 2019, and then kept moving southwestward in Sanming City between 2019 and 2021.

3.4. Driving Factors of TEE in the GFZ Coastal City Clusters

The TEE divergence is the result of the combined effect of multiple factors, and this study uses a geographic detector to detect the factor influence on the TEE divergence of the GFZ coastal city clusters using six factors. Considering the limited number of cities and data availability in the GFZ coastal city clusters, the calculation by year was more sensitive than the full time period and could avoid the long-cycle smoothing effect masking the short-term abrupt changes; therefore, five years, namely, 2010, 2013, 2016, 2019, and 2021, were selected for the analysis of the influencing factors [34]. The q-value of the factor detector results characterizes the magnitude of the factor’s explanatory power of a factor on spatial divergence (

Table 6. Results of the factor detector (q-value).

Impact Factor	2010		2013		2016		2019		2021
	q	Sorting	q	Sorting	q	Sorting	q	Sorting	q	Sorting
Economic development (X1)	0.435	5	0.354	4	0.239	5	0.252	3	0.208	1
Tourism industrial structure (X2)	0.085	6	0.126	6	0.13	6	0.037	6	0.104	6
Transportation (X3)	0.472	2	0.341	5	0.3	4	0.123	4	0.2	3
Government intervention (fiscal expenditure) (X4)	0,471	3	0.573	2	0.509	1	0.555	1	0.204	2
Urbanization level (X5)	0.496	1	0.689	1	0.366	3	0.322	2	0.178	4
Green space level (X6)	0.595	4	0.482	3	0.395	2	0.116	5	0.125	5

).

In 2010, the contribution rate of each influence factor was in the following order: urbanization level > transport > fiscal expenditure > green land level > economic development > tourism industry structure; in 2013: urbanization level > fiscal expenditure > green space level > economic development > transportation > tourism industry structure; in 2016: fiscal expenditure > green land level > urbanization level > transport > economic development > tourism industry structure; in 2019: fiscal expenditure > urbanization level > economic development > transportation > green land level > tourism industry structure; and in 2021: economic development > fiscal expenditure > transportation > urbanization level > green land level > tourism industry structure. Among the influencing factors in 2010–2021, the contribution of fiscal expenditure and urbanization level to the TEE was the largest, and their explanatory strength was relatively stable.

The interaction detector in the geographical detector was used to further analyze the interaction of the factors affecting TEE in the GFZ coastal city clusters by adopting the third stage of eco-efficiency data; adding two factors (traffic and urbanization level); and according to the relevant studies [46] and the actual situation of eco-tourism in Guangdong, Fujian, and Zhejiang, the interactions between the six influencing factors. For the four years (2010, 2016, 2019, 2021) of interaction used in the analysis, the results are shown in Figure 5. Any two factors of interaction between the two factors are presented as two-factor enhancement or nonlinear enhancement of the characteristics of the situation, not observed independent of each other, or as a weakening of the situation. This indicates that the interaction of any two factors significantly affects TEE more than the effect of a single factor. In other words, the evolution of TEE in the GFZ coastal city clusters is the result of the joint action of multiple influencing factors.

The study of interaction detection revealed that the interaction between the influencing factors significantly outweighed the influence of a single factor on the TEE. Specifically, the interaction between economic development and industrial structure was the most significant in 2010; the interaction between economic development and government intervention was stronger in 2013; in 2016, the interaction between the external environment was significantly stronger than the other influencing factors, indicating that the dominance of external factors was more prominent at that stage; in 2019, there was a strong interaction between the urbanization level and the greening level; and by 2021, the interaction between transport and government intervention also reached the highest level. The above results indicate that the role of external development factors increased in the development stage of the GFZ coastal city clusters in 2016–2021, suggesting that tourism relies mainly on the accumulation of internal resources at the initial stage, but as its development process progresses, synergistic development with the integration of external resources becomes increasingly important.

4. Discussions

4.1. TEE Measurement Based on the Three-Stage Super-Efficient SBM Approach

Employing the enhanced three-stage super-efficient SBM approach, this work quantified the TEE of the GFZ coastal city clusters. This model boasts significant methodological advantages. Xie et al. also used the three-stage super-efficient DEA model to measure the industrial environmental efficiency in China [48]. However, the model exhibits an inherent flaw in that it is incapable of handling input–output slack variables, a deficiency that can result in the overestimation of efficiency values. Li improved it to a three-stage super-SBM to optimize the slack variable problem [49]. Yizhen et al. assessed the green innovation efficiency (GIE) of energy-intensive firms through a three-stage super-efficient SBM approach, measuring a more accurate GIE. The GIE level of listed energy-intensive enterprises in China is reflected by this model [22].

Based on the evolution and advantages of the above methods, this study adopts the improved three-stage super-efficient SBM approach to effectively compensate for the limitations of the three-stage super-efficient DEA model by optimizing the relaxation variables so that the efficiency value can be calculated more accurately. This methodological improvement enables this study to achieve a more accurate measurement of TEE for the GFZ coastal city clusters while eliminating the interference from the external environment. Specifically, in this study, first, under the perspective of adopting the three-stage super-efficient SBM approach, the TEE of the GFZ coastal city clusters in the third stage was higher than that of the first stage, which indicates that the GFZ coastal city clusters was underestimated without taking into account the external environment and random disturbances, and therefore, the third-stage TEE was more accurate. As early as 2002, Fried proved in his theoretical and empirical research that if the value of efficiency in the third stage is higher than in the first stage, it usually means that the first-stage efficiency is underestimated. This underestimation stems from the fact that the efficiency value of the first stage is mixed with the negative effects of unfavorable external environmental factors or bad stochastic disturbances [34]. Thus, the third-stage efficiency value is considered to be a more accurate reflection of the “true” management efficiency of the urban agglomerations after removing these external disturbances [34]. Conversely, the results show that the input–output ratios in the coastal metropolitan agglomerations of Zhejiang, Fujian, and Guangdong had an unequal connection; inefficiency is indicated by a TEE of less than 1 (100 percent). This is because under the DEA framework (either CCR, the BCC model, or the SBM approach), the root cause of non-efficiency is disproportionate regarding the input–output ratio [50]. Second, the average TEE of each prefecture-level city after adjustment reached a level above the medium level, which suggests that the TEE inefficiency of each prefecture-level city before the adjustment was not entirely due to its own managerial inefficiency and was mainly related to relatively disadvantaged external environmental variables [34].

4.2. The Spatiotemporal Evolution of TEE

TEE serves as the crucial metric for assessing the relationship between regional tourist economy development and ecosystem conservation, which is essential for tourism’s sustainable development. The GFZ coastal city clusters is of particular interest in the context of China’s cross-provincial economic collaboration and international engagement since it is highly concentrated in energy consumption and pollution emissions, making them key practice areas for green tourism development. By analyzing the TEE time-series change in the GFZ coastal city clusters, distinct trajectories of TEE variation were observed across prefectures and cities within these coastal clusters during 2020–2021, and the whole showed a fluctuating upward trend, which suggests that ecological protection of the Guangdong, Fujian, and Zhejiang coastal urban agglomeration during the study period achieved a certain degree of success.

Among the cities, Putian City, Sanming City, Nanping City, and Lishui City showed a decreasing trend in the TEE, which may have been related to the development of industrialization; Putian, Sanming and Nanping have higher industrial development in Fujian Province, which produces more sources of pollution, and there is a phenomenon of over-consumption of resources in the pursuit of economic development. Furthermore, there are more mountainous regions in Fujian Province compared with Guangdong Province and Zhejiang Province, which have a more complex topography and are not convenient for transportation. Fu et al. examined the contribution of natural environmental factors to TEE and found that natural environmental factors had a greater effect on TEE [40]. This confirms that in the study region, the increase in the TEE over time in the more mountainous prefecture-level cities in Fujian Province was not significant and had a slight downward trend, which is in line with the research expectations.

The TEE of the GFZ coastal city clusters spatially showed a top-down decreasing trend in the early period and reversed to a top-down increasing trend in the later period from 2010 to 2021. Specifically, there was a significant increase in the high-value areas and a significant decrease in the middle- and low-value areas, which are indications that the TEE of the GFZ coastal city clusters was heading in the right direction. It is worth noting that Lishui City’s declining efficiency values from 2013 to 2021 were associated with a rebound in sulfur dioxide emissions and a decline in desired output, suggesting that the city’s efficiency was constrained by internal management rather than the external environment, validating the core assertion of endogenous growth theory, i.e., institutional quality determines the factor allocation efficiency [51].

4.3. The Driving Mechanism of TEE

Further analysis showed that the drivers of the overall increase in regional TEE are compounded, with productivity gains benefiting significantly from green technological advances regarding productivity per unit of labor, and environmental regulations further driving investment in green technology innovation through the “push mechanism” created by external pressures [52]. In particular, the overarching roadmap inaugurated a new development stage, where China’s comprehensive national strategy for economic and social advancement over the 2021–2025 period, alongside its long-term objectives through to 2035, points out that innovation-driven development has a central position in the overall situation of China’s modernization and that advancing a comprehensive green transformation of economic and social development is an important path to building a beautiful China. Policies have guided local governments to pay increasing attention to green development, which has led to continuous improvement of the TEE. The TEE gradually spread in the northeast–southwest direction, with the center of gravity volatility shifting to the southwest. This may have been related to the technology spillover and policy response lag; the study area belonged to the eastern coastal cities, and there is a time lag in the diffusion of its green technology to central and western provinces, resulting in the center of gravity migration path being synchronized with the policy implementation cycle. At the same time, it shows that the latecomer advantage of central and western China has emerged, and inland areas, such as Sanming City, have accelerated their efficiency after 2016 by undertaking industrial transfers, pulling the center of gravity to the southwest [53].

Through the factor detector and interaction detection, it was found that the influencing factors have explanatory power on TEE, and the interaction between the factors significantly outweighs the influence of a single factor on tourism eco-efficiency. Examined from the perspective of dynamic evolution, different historical stages present differentiated dominant driving factors and their interaction mechanisms.

Specifically, the first stage (2010) manifested the double-wheel drive of economic scale and industrial structure, with the tourism industry relying on crude growth in the early stage, and the efficiency enhancement mainly relied on the expansion of economic scale and optimization of industrial structure [47]. In the second stage (2013), the interaction between economic development and government intervention showed that the study area’s government was able to enhance the eco-efficiency of tourism through policies (e.g., stimulus plans) [54]. The third stage (2016) highlighted the constraints of the external environmental factors, and with the tightening of environmental regulations and the risk of climate change, the traditional development model is facing double pressure: on the one hand, the internalization of environmental costs has forced enterprises to improve their production processes; on the other hand, the frequent occurrence of extreme weather events has prompted the industry to restructure its risk management system. This shift confirms the theoretical expectation of Porter’s hypothesis, which suggests that stringent environmental regulations may induce an innovation compensation effect, forcing the tourism industry to pay attention to external environmental impacts. The fourth stage (2019) was characterized by a strong interaction of spatial elements, which stemmed from the compression of ecological space due to rapid urbanization and forced a synergy between “green urbanization” policies and ecological restoration of the tourism industry [55]. The fifth stage (2021) demonstrated the synergistic innovation of the system and technology; the recovery of tourism in the post-epidemic era relies on the transport network’s resilience and the government’s precise regulation, and transport has an important supporting role for sustainable development [56]. Therefore, policies need to focus on multifactor linkage effects as tourism moves from reliance on internal resources to external synergistic development. This finding will also provide an empirical reference point for subsequent policymakers’ decisions in the region.

It is worth noting that Guangdong, Fujian, and Zhejiang Provinces have differences in environmental policy intensity and financial support, but in the study, the urban agglomeration was considered as a homogeneous whole, and the impact of inter-provincial policy heterogeneity on the TEE was not analyzed. In addition, five nodal years, namely, 2010, 2013, 2016, 2019, and 2021, were selected for the driver analysis to represent the different stages, but a single year may be subject to short-term fluctuations, which may limit the impact effect.

4.4. Implications and Outlooks

Based on the aforementioned results, this paper proposes the following policy implications:

• It is recommended that the municipal government establish a “tourism eco-efficiency synergistic enhancement center”; delineate the ecological red line of tourism development; implement a system of replacing industrial land with tourism land for cities with declining efficiency, such as Putian and Sanming; set up a subsidy fund for green technology; and prioritize support for the introduction of sulfur dioxide abatement technology for endogenous efficiency shortfalls, such as in Lishui.
• Although this study did not analyze policy differences between provinces, the spatial and temporal distribution characteristics of ecological governance suggest that differences in environmental regulatory intensity between Guangdong, Fujian, and Zhejiang Provinces should be considered. These differences have led to variations in TEE growth rates. For example, Fujian Province has established an inter-city ecological compensation mechanism, transferring funds for environmental governance from efficient cities, such as Ningde and Fuzhou, to less developed cities, like Sanming and Nanping; Zhejiang Province has implemented digital regulation, integrating sulfur dioxide emission data from Lishui City into the provincial ecological cloud platform for real-time tracking; the technological advantages of the Bay Area will be enhanced in Guangdong Province; and the “cultural tourism–industrial symbiosis park” model will be promoted in Chaozhou and Shantou to achieve a balance between development and protection.

This study has some limitations and uncertainties that need to be improved in future research. First, the study did not analyze the GFZ coastal city clusters as a heterogeneous unit by considering the inter-provincial policy differences in depth. Second, the potential endogeneity between certain influencing factors (e.g., economic development and government intervention) and the TEE was not addressed through the instrumental variable approach or lagged variable analysis, which is a direction that can be further discussed in subsequent studies. Third, future studies can consider a sensitivity analysis of the data to test the robustness of the data model.

5. Conclusions

This study involved the use of a sophisticated measurement model (three-Stage super-efficient SBM approach) to accurately assess the TEE of the GFZ coastal city clusters from 2010 to 2021. The study area’s spatiotemporal characteristics and influencing factors were then analyzed through the application of standard deviation ellipsoids and geographic probes. The main conclusions are as follows:

• This study is based on a three-stage super-efficiency SBM approach, and after accounting for the role of extraneous environmental variables, including regional affluence, sectoral composition, government intervention, and urban greening, the mean value of TEE for each municipality in the study area was significantly higher, and external variables had a significant effect on the efficiency (Gamma value > 0.6).
• Regarding the temporal evolution, the TEE as a whole showed a fluctuating upward trend; spatially, the TEE within the GFZ coastal city clusters showed a top-down downward trend from 2010 to 2016 and a top-down upward trend from 2016 to 2021. The directional distribution ellipse for TEE exhibited a northeast–southwest orientation, with its center of gravity displaced southwestward.
• Regarding the TEE influencing factors, TEE was mainly influenced by internal drivers in 2010–2016, and the role of external development factors within the Guangdong-Fujian–Zhejiang coastal metropolitan belts increased in 2016–2021, indicating that the tourism industry relied mainly on the accumulation of internal resources in the initial period, and that synergistic development integrating with external resources became increasingly important over time.