Measurement and Convergence Analysis of the Green Total Factor Productivity of Citrus in China

Abstract

Drawing on panel data for eight major citrus-producing provinces in China from 2008 to 2021, this study employs the super-efficiency EBM model—which incorporates both radial proportion and non-radial slack variables—to measure citrus green total factor productivity (GTFP). Temporal changes are investigated via the GML index, while regional disparities and convergence patterns are examined through a series of complementary techniques, thereby offering a comprehensive view of the sector’s green and coordinated development. The results reveal that, from a static perspective, the technical efficiency of most citrus-producing provinces remains below the production frontier. Dynamically, regional GTFP diverged markedly over the study period, with technical efficiency serving as the principal driver of growth. Convergence tests show no evidence of σ-convergence for the nation as a whole or for any of the three major producing regions. Absolute and conditional β-convergence coexist at the national level and in the upper–middle Yangtze region; the Zhejiang–Fujian hills exhibit no β-convergence, whereas the Guangdong–Guangxi hills display conditional β-convergence only. The findings indicate substantial room for improvement in China’s citrus GTFP. We therefore recommend that each region (1) accelerates green-technology innovation, (2) designs differentiated yet coordinated regional strategies, (3) institutionalizes long-term safeguards for green development, and (4) deepens international cooperation to enhance global competitiveness.

1. Introduction

Citrus is one of the most widely planted agricultural cash crops in southern China [1]. In 2008, China’s citrus planting area and output exceeded those of Brazil for the first time, making it the world’s largest citrus grower and producer [2]. In 2021, China’s citrus planting area reached 2.9227 million hectares, accounting for 22.82% of the country’s total fruit planting area. The output was 55.956 million tons, which accounted for 18.67% of the national fruit output and ranked first in the national total fruit production [3]. As the world’s largest citrus producer, China’s citrus industry plays an important role in boosting farmers’ income and rural economic development. However, China’s citrus sector has been locked into an input- and resource-intensive production paradigm historically. This has triggered increasingly severe problems of resource waste, environmental contamination, and ecological degradation. Fertilizer use offers a salient illustration: survey data reveal that excessive application of nitrogen, phosphorus, and potassium occurs on 57.3%, 76.6%, and 69.1% of citrus orchards, respectively, while the corresponding nutrient-use efficiencies remain between 20% and 30%. The resultant economic inefficiency is compounded by soil salinization, water eutrophication, and elevated greenhouse gas emissions [4,5]. In December 2023, the Central Rural Work Conference of China emphasized the need to stick to the red line of cultivated land. Against the backdrop of increasingly tight constraints on cultivated land resources and increasing pressure on the ecological environment, the traditional scale-expansion growth model has struggled to continue promoting the development of the citrus industry. Furthermore, intensive application of chemical fertilizers and pesticides in citrus cultivation systems poses significant environmental and public health concerns, including ecological degradation, food safety contamination, and adverse health impacts [6].

The 20th National Congress report underscores the strategic imperative of enhancing total factor productivity (TFP) as a cornerstone for achieving high-quality development. Within this framework, advancing agricultural TFP emerges as a pivotal component in accelerating the modernization of China’s agricultural sector and realizing the national objective of agricultural power construction. Agricultural green total factor productivity refers to the realization of agricultural production goals while minimizing energy consumption, emissions, and resource waste, so as to achieve the synergistic development of agricultural growth and environmental protection [7]. Against the backdrop of global sustainable agriculture imperatives, agricultural GTFP advancement is now recognized as fundamental to balanced economic–environmental–social progress. In this context, the improvement of green total factor productivity in the citrus industry, that is, reducing the excessive input of factors such as energy, chemical fertilizers and pesticides and reducing environmental pollution while ensuring yield and economic benefits, has become an inevitable choice to break the resource and environmental constraints of the citrus industry and achieve sustainable development, and is also an important way to promote the high-quality development of the citrus industry.

Although existing studies have preliminarily discussed the green total factor productivity of citrus, there is still room for further deepening in terms of measurement methods, index system construction, and regional difference analysis. Specifically, it is not clear which factors can effectively promote the improvement of China’s citrus green total factor productivity, and whether China’s citrus industry has achieved regional coordinated development. To address these research gaps, this study employs panel data (2008–2021) from China’s eight primary citrus-producing provinces. We construct a novel measurement framework that incorporates both desirable outputs (economic and ecological benefits) and undesirable outputs (environmental pollutants). Utilizing the super-efficiency EBM model combined with the Global Malmquist–Luenberger (GML) index, we conduct a dual-aspect evaluation of citrus green total factor productivity (GTFP), examining both static efficiency levels and dynamic evolutionary patterns. This study further investigates the spatiotemporal evolution patterns and regional heterogeneity of citrus GTFP, while conducting rigorous convergence analysis. It is hoped that this study will broaden the scope of methods for measuring agricultural productivity through these multidimensional tests, and provide theoretical insights and policy implications for promoting the sustainable development of the citrus industry.

2. Literature Review

2.1. Studies on Green Total Factor Productivity in Agriculture

The advent of the agricultural green revolution has spurred growing scholarly interest in green total factor productivity (GTFP) measurement. Current academic investigations primarily concentrate on three core dimensions: (1) development of comprehensive evaluation indicator systems, (2) advancement of measurement methodologies, and (3) examination of determinant factors affecting productivity performance.

2.1.1. Indicator System Construction for Agricultural Green Total Factor Productivity

Diverging from traditional TFP metrics, agricultural green total factor productivity (GTFP) evaluation internalizes negative environmental outputs (e.g., CO₂ equivalents, nitrogen leaching) through directional distance functions [8,9], enabling a comprehensive assessment of sustainable intensification potential while maintaining production efficiency. The construction principles of the indicator system are as follows: first, input factors (e.g., labor and machinery) and output indicators (e.g., total agricultural output value and carbon emissions) are processed separately to ensure the inclusion of all key variables; second, the combination of dynamic and static methods: the directional distance function (DDF) is combined with super-efficiency models (such as SBM), which can not only statically evaluate efficiency but also dynamically track technological progress [10,11]. In terms of the selection of measurement indicators, scholars usually choose gross agricultural product or value added of agricultural production [12,13] as the desired output of agriculture, and agricultural carbon emission [14] as the undesired output. Yang et al. [15] and Yao et al. [16] advanced agricultural productivity measurement by incorporating carbon sequestration valuation into their analytical framework, where the gross agricultural output value served as the primary desirable output while accounting for ecosystem services, and Li et al. [17] took into account agricultural surface source pollution when taking agricultural carbon emission as the undesired output. In terms of input factors, scholars [18,19] mainly used indicators such as the number of people employed in the primary industry, the total sown area of crops, the total power of machinery, and the amount of fertilizer applied.

2.1.2. Methodological Approaches for Agricultural Green TFP Measurement

The measurement of agricultural green total factor productivity (GTFP) primarily employs two methodological approaches: parametric and non-parametric techniques. Within parametric methods, stochastic frontier analysis (SFA) has been widely adopted since its foundational development by Aigner et al. and Meeusen et al. SFA estimates green production efficiency by specifying a functional form for the production process and distinguishing inefficiency from stochastic noise [20,21]. In contrast, non-parametric methods predominantly rely on data envelopment analysis (DEA) and its extended models, which evaluate the relative efficiency of decision-making units by constructing an empirical production frontier. Notably, Chung et al. advanced this approach by integrating undesirable outputs, such as pollution emissions, into a directional distance function framework, as demonstrated in their seminal study of Swedish pulp mills. This dual methodological paradigm—encompassing both SFA’s econometric rigor and DEA’s flexibility in handling undesirable outputs—provides comprehensive tools for assessing sustainable agricultural productivity [22]. To mitigate the measurement errors that conventional DEA models may incur when undesirable outputs are present, Tone and Tsutsui (2010) simultaneously incorporated both the radial ratio between frontier and observed inputs and the non-radial, input-specific slacks into the efficiency-measurement framework, thereby developing the Epsilon-Based Measure (EBM) model [23]. Subsequently, to overcome the classical DEA limitation of being unable to differentiate among decision-making units whose efficiency scores equal unity, Andersen and Petersen (1993) introduced the super-efficiency concept [24]. By integrating these two strands, the Super-EBM model retains the EBM model’s dual capability of accounting for both radial and non-radial slacks while further enabling the discrimination and ranking of efficient units. Peng et al. used the super-efficiency EBM model to measure the inclusive green efficiency of cities [25]. The Malmquist–Luenberger index extends the application of the DEA methodology in intertemporal dynamic analysis. Myeki et al. used the ML index to measure agricultural efficiency in the 2000–2019 period, and the growth of agricultural green total factor productivity (GTFP) in 49 African countries [26]. To more accurately reflect the changes in long-term production efficiency, Oh combined the concept of global Malmquist productivity with the directional distance function and proposed the GML index [27]. Zhu et al. used GML to quantify the productivity changes in different regions of China at different time periods when studying China’s agricultural green total factor productivity, and analyzed the roles of technical efficiency and technological progress [28]. In summary, the differences between the Traditional DEA/Malmquist Index and the Super-EBM + GML Index are summarized in

Table 1. Method comparison and explanation.

Distance Function	Traditional DEA/Malmquist Index	Super-EBM + GML Index
Distance function	Purely radial adjustment → upward bias in efficiency scores	Radial + non-radial slacks → unbiased “true” efficiency
Undesirable outputs	Ignored → environmental costs omitted	Explicitly incorporates CO2 emissions and non-point-source pollution → green efficiency
Frontier	Contemporaneous frontier → intertemporal distortion	Global frontier → stable intertemporal comparison
Efficiency > 1	Cannot discriminate → frontier crowding	Super-efficiency ranking → further differentiation of leading provinces

:

2.1.3. Factors Influencing Green Total Factor Productivity in Agriculture

Studies examining the determinants of agricultural green total factor productivity (GTFP) have identified multiple influential factors, including environmental regulation, demographic composition, mechanization intensity, and industrial agglomeration. Empirical research by Xu et al. demonstrates this relationship through their analysis of Chinese agricultural systems, revealing that enhanced environmental regulations significantly stimulate technological innovation in the sector [29]. Regarding the impact of demographic structure, different scholars have different views. Jin et al. studied the impact of population aging on China’s agricultural total factor productivity, and the results show that population aging can significantly enhance agricultural green total factor productivity, and there are obvious regional differences in the impact, in which the role of the western and the primary grain-producing areas is more significant [30]. Deng et al. [31] found a negative correlation between rural population composition and GTFP performance, revealing that certain demographic indicators may limit productivity growth. As a supplement to this discovery, Wang et al. [32] demonstrated the limited ability of traditional agricultural mechanization to promote green technological progress, emphasizing the crucial role of specially designed green mechanization technologies in improving sustainable productivity. Han et al. [33] further elucidated the spatial dimension of GTFP dynamics, and their spatial Durbin model analysis confirmed that agricultural industry agglomeration can generate both local productivity benefits and positive spatial spillover effects on neighboring regions. On the basis of these basic studies, subsequent research has advanced the field through three key analytical extensions: firstly, Liu et al. incorporated spatiotemporal evolution patterns into their research [34], secondly, Liu et al. conducted interregional difference decomposition [35], and Ji et al. proposed convergence trend analysis [36], jointly establishing a comprehensive research paradigm to understand the dynamics of agricultural GTFP.

It is worth noting that climate change, as a crucial external driver of agricultural green total factor productivity (GTFP), has been increasingly incorporated into analytical frameworks in recent years. Using global panel data on crop yields, Wing et al. (2021) find that climate warming may reduce the output of cash crops such as citrus by 3–12% through heat stress and altered precipitation patterns [37]. Complementarily, Shmelev et al. (2021), in a study of Kazakhstan, confirm that both El Niño–Southern Oscillation (ENSO) events and soil-moisture deficits significantly suppress crop yields [38].

2.2. Studies on the Efficiency of Citrus Production

Existing literature on citrus production efficiency in China has concentrated primarily on measuring technical efficiency and total factor productivity, and has proceeded along two analytical dimensions—conventional input factors and ecosystem services—thereby offering a diversified set of perspectives and findings that deepen our understanding of citrus production efficiency.

2.2.1. Studies on Citrus Production Efficiency from the Perspective of Traditional Input Factors

A substantial body of literature has employed diverse methodological approaches to examine citrus production efficiency in China, centering on how traditional inputs—namely labor, chemical fertilizers, and machinery—affect citrus green total factor productivity (GTFP). Li et al. [39] used the Malmquist index method to measure the total factor productivity of citrus in China’s main citrus-producing provinces from 2008 to 2014, and found that there was a large interprovincial gap in total factor productivity of citrus; Xu et al. [40] used DEA to measure the input–output indicators of China’s seven primary citrus-producing regions from 2009 to 2015, and found that technological progress was the major influencing factor of total factor productivity improvement of China’s citrus in the study period; Fang et al. [41] measured citrus total factor productivity based on panel data of China’s citrus production from 2011 to 2016, and similarly concluded that technological progress is the primary factor affecting total factor productivity in the citrus industry, and found that China’s citrus total factor productivity has been on a downward trend since 2011; Lin et al. [42] measured surface source pollution constraints under the 2008–2017 period of China’s citrus production efficiency, and analyzed regional differences, and found that citrus production efficiency was higher in central and eastern China, and lower in southern and southwestern China, where the main reasons for the loss of citrus production efficiency were the high total nitrogen and total phosphorus losses and the redundancy of fertilizer inputs.

2.2.2. Studies on Citrus Production Efficiency from the Perspective of Ecosystem Services

Citrus production is not only contingent on conventional inputs but also deeply embedded within a suite of ecosystem services, such as pollination, water regulation, and climate moderation. Shmelev et al. (2023), in a multidimensional assessment conducted in France, demonstrate that the sustainability of agro-systems—including orchards—critically depends on pollinator populations (e.g., Apis spp.) and stable hydrological cycles [43]. In China, key citrus regions such as the upper–middle Yangtze basin and the Zhejiang–Fujian hills have experienced measurable declines in pollination services, largely attributable to habitat fragmentation (IPBES, 2019) [44]. Extreme drought events—exemplified by the 2022 Yangtze River drought—have further threatened irrigation water availability. Moreover, the global ecosystem typology advanced by Keith et al. (2022) classifies citrus orchards as “managed terrestrial systems,” whose functional stability is shaped jointly by anthropogenic management and underlying ecological processes [45]. Against this backdrop, future research must quantify the degree to which citrus production depends on ecosystem services (ES), for instance by evaluating how pollinator diversity or watershed retention capacity contributes to citrus green total factor productivity.

2.3. Summary

Previous studies have provided valuable benchmarks for both indicators and methodologies, yet three avenues for refinement remain evident. (i) Research scope: The extant literature predominantly examines the green total factor productivity (GTFP) of apples [46], wheat [47], vegetables [8], and other crops, whereas only a handful of papers preliminarily estimate GTFP for Chinese citrus. These studies, however, rely on outdated sample periods and therefore cannot reflect recent shifts in ecological and socio-economic conditions. (ii) Methodological limitations: Existing citrus studies almost exclusively adopt the conventional DEA-Malmquist framework, which is incapable of simultaneously handling radial proportions and non-radial slacks, suffers from efficiency scores “clustering at unity,” and rarely incorporates undesirable outputs such as carbon emissions and non-point-source pollution. Consequently, published estimates fail to capture environmental benefits. Moreover, convergence analyses of citrus GTFP are virtually absent; whether regional disparities are narrowing toward coordinated development has yet to be empirically verified. (iii) Indicator narrowness: Prior work focuses on economic outputs—yield, output value, etc.—while neglecting the ecological benefits intrinsic to citrus cultivation.

To address these gaps, this paper offers three extensions. First, the research object is updated: we center on citrus GTFP and extend the sample window to 2008–2021. Second, methodologically, we integrate Super-EBM with the GML index to overcome the discriminative limitations of traditional DEA and to capture intertemporal technological progress and efficiency gains. We further investigate σ- and β-convergence of citrus GTFP at the national level and across the three major producing regions, thereby elucidating regional convergence mechanisms and divergent pathways of green development. Third, the indicator system is broadened: desirable outputs now include ecological benefits, while undesirable outputs such as environmental pollutants are explicitly incorporated into the GTFP measurement framework.

These refinements provide a richer perspective on citrus GTFP and furnish more nuanced empirical evidence for differentiated green-transition policies. Future research could embed citrus production within a broader ecosystem-services framework, dynamically assessing how pollinator diversity, water-retention capacity, and extreme-climate events influence industry-level GTFP, and quantifying the long-term risks and adaptation pathways induced by ecological change. Such efforts would offer forward-looking scientific guidance for enhancing the resilience of the citrus sector and optimizing its spatial configuration under a changing climate.

3. Research Methods

3.1. Super-Efficiency EBM Model

Citrus production involves both radial and non-radial features in its inputs and outputs, and China’s citrus cultivation is characterized by pronounced environmental externalities. Neither the traditional radial DEA nor the non-radial SBM model can therefore depict the production process accurately; moreover, these models cannot differentiate among decision-making units whose efficiency scores exceed unity when multiple inputs and outputs are present [48]. In contrast, the EBM model simultaneously accommodates radial proportions and non-radial slacks, yielding unbiased optimal efficiency scores, whereas the super-efficiency EBM variant further enables the effective ranking of units with efficiency values greater than one [49]. Accordingly, this study employs the super-efficiency EBM model—constructed by integrating Andersen et al.’s [24] super-efficiency concept with Tone’s [23] EBM framework—to estimate citrus green total factor productivity. The corresponding computational formula is presented as Equation (1):

$$\begin{gathered} K^{*}=\underset{\theta ,\phi ,\gamma ,s^{-},s^{+}}{min}{\displaystyle \frac{\theta +{\epsilon }_x\sum _{i=1}^m\frac{w_i^{-}{s}_i^{-}}{x_{io}}}{\phi -{\epsilon }_y\sum _{r=1}^s\frac{w_r^{+}{s}_r^{+}}{y_{ro}}-{\epsilon }_b\sum _{q=1}^p\frac{w_q^{b-}{s}_q^{b-}}{b_{qo}}}} \\ s.t.\left\{\begin{array}{c}{\displaystyle \sum _{j=1,j\ne o}^n}{x}_{ij}^t{\gamma }_j^t-s_i^{-}\le \theta x_{io} i=\mathrm{1,2},\dots ,m\\ {\displaystyle \sum _{j=1,j\ne o}^n}{y}_{rj}^t{\gamma }_j^t+s_r^{+}\ge \phi y_{ro} r=\mathrm{1,2},\dots ,q\\ {\displaystyle \sum _{j=1,j\ne o}^n}{b}_{qj}^t{\gamma }_j^t-s_q^{b-}\le b_{qo} q=\mathrm{1,2},\dots ,p\\ {\displaystyle \sum _{j=1,j\ne o}^n}{\gamma }_j^t=1\\ \gamma \ge 0,s_i^{-}\ge 0,s_r^{+}\ge 0,s_q^{b-}\ge 0\end{array}\right. \end{gathered}$$

(1)

In Equation (1): K* denotes the citrus production efficiency of the province under evaluation. j indexes all possible decision-making units (DUM), each ${DMU}_j$ consumes m inputs $x_{ij}$ (i = 1, 2,…, m), produces s desirable outputs $x_{rj}$ (r = 1, 2,…, s), and p undesirable outputs $b_{qj}$ (q = 1, 2,…, p); γ is the linear-combination weight attached to ${DMU}_j,$ $x_{io}$, $y_{ro}$ and $b_{qo}$ are, respectively, the input, desirable output, and undesirable output vectors of the evaluated ${DMU}_o$, $s_i^{-}$, $s_r^{+}$ and $s_q^{b-}$ are the slack variables for the i-th input, r-th desirable output, and q-th undesirable output; $w_i^{-}$, $w_r^{+}$ and $w_q^{b-}$ denote the weights assigned to the ith input, the rth desirable output, and the qth undesirable output, respectively. In this study, all weights were set to 0.5 during the computation; θ and φ are the radial planning parameters; $\epsilon$ ∈ [0, 1] is the key non-radial weight that reflects the importance of the slacks in the efficiency score; K* is the optimal efficiency under variable returns to scale. A value K* < 1 indicates inefficiency; K* ≥ 1 indicates efficiency, with larger K* implying higher efficiency. The specification shown in Equation (1) incorporates undesirable outputs; when they are excluded, ${\epsilon }_b$ is set to zero. Following Wang (2023) [32], we employ MaxDEA 9 to derive the weights, thereby objectively reflecting the relative importance of each input and output and avoiding subjective weight assignment [48].

In the empirical specification, the model is estimated under both constant returns to scale (CRS) and variable returns to scale (VRS). Under CRS, the assessment focuses solely on technical efficiency (TE) while abstracting from scale effects; the resulting score is the conventional technical-efficiency index, which can be decomposed into pure technical efficiency (PTE) and scale efficiency (SE). Under VRS, the evaluation simultaneously accounts for technical and scale efficiencies and yields the pure technical-efficiency measure (PTE). Equation (1) is formulated under the VRS assumption; removing the constraint $\sum _{j=1,j\ne o}^n{\gamma }_j^t=1$ yields the CRS specification. Scale efficiency is then obtained as SE = TE ÷ PTE, providing an indicator of whether the DMU operates at its optimal scale. When undesirable outputs are incorporated, the corresponding concepts are labeled environmental technical efficiency (GTE), environmental pure technical efficiency (GPTE), and environmental scale efficiency (GSE).

3.2. Global Malmquist–Luenberger Index

Considering that the efficiency values estimated by the above super-efficient EBM model are static and cannot be compared across periods, at the same time, the traditional ML index may have the problem of no feasible solution when measuring the hybrid directional distance function. Therefore, this paper draws on the study of Bai et al. [50] and adopts the Global Malmquist–Luenberger (GML) index, which is further decomposed into technical efficiency change and technical progress change, in order to analyze the intertemporal change of total factor productivity in citrus green, which is computed as shown in Equation (2):

$$\begin{array}{ll}\hfill {GML}^{t,t+1}& ={\displaystyle \frac{1+D_t\left(x^t,y^t,b^t\right)}{1+D_{t+1}\left(x^{t+1},y^{t+1},b^{t+1}\right)}}\\ & \times \left[{\displaystyle \frac{\frac{1+D_t^G\left(x^t,y^t,b^t\right)}{1+D_t\left(x^t,y^t,b^t\right)}}{\frac{1+D_{t+1}^G\left(x^{t+1},y^{t+1},b^{t+1}\right)}{1+D_{t+1}\left(x^{t+1},y^{t+1},b^{t+1}\right)}}}\right]\\ & ={GEC}^{t,t+1}\times {GTC}^{t,t+1}\end{array}$$

(2)

The GML index in Equation (2), denoted as GML^t,t+1, quantifies the dynamic change in citrus green total factor productivity (GTFP) across consecutive periods (t to t + 1) at the provincial level. $D_t^G\left(x^t,y^t,b^t\right)$ is the direction vector of the reference set, where $\mathrm{D}$ is the directional distance function. GEC is the technical efficiency index, and GTC is the technological progress index. Values exceeding unity (GML > 1) signify progressive enhancement of green total factor productivity, reflecting positive technological or efficiency advancements. Conversely, values below unity (GML < 1) denote regressive productivity performance, indicating deteriorations in production efficiency or environmental performance.

Since the GML model measures the change in green total factor productivity, in order to obtain the actual value of green total factor productivity (GTFP) of citrus in China, this paper draws on Zhu’s calculation method [51]. The specific method is as follows: assuming that the GTFP in 2008 is 1, the GTFP in 2009 is the GTFP in 2008 multiplied by the GML index in 2008–2009, and so on. In accordance with the distinct theoretical frameworks of constant returns to scale (CRS) and variable returns to scale (VRS), this study employs separate computational procedures for the technical efficiency change index (GEC) and technological progress index (GTC). The methodological approach for these calculations maintains consistency with the estimation framework of the Global Malmquist–Luenberger (GML) index, ensuring comparability across productivity components. Ultimately, the citrus GTFP and its decomposed GEC and GTC for the eight main citrus-producing provinces in China from 2008 to 2021 can be obtained. The calculation formula is shown in Equation (3):

$${GTFP}_{t+1}={\mathrm{G}\mathrm{M}\mathrm{L}}^{t,t+1}\times {GTFP}_t$$

(3)

When undesired output is not taken into account, the GML, GEC, and GTC are equivalent to the ML (total factor productivity), TC (technological progress index), and EC (technical efficiency index), respectively.

3.3. Convergence Test

The concept of σ-convergence characterizes the temporal reduction in cross-regional disparities of citrus green development levels. Following the methodological approach of Ma et al. [52], this study employs the coefficient of variation to empirically test for σ-convergence in citrus green total factor productivity (GTFP). The formal specification of this measurement is presented in Equation (4):

$${\alpha }_t=\frac{\sqrt{\frac{1}{n}\sum _{i=1}^n{\left(x_{i,t}-{\overline{x}}_t\right)}^2}}{{\overline{x}}_t}$$

(4)

In Equation (4), ${\alpha }_t$ is the coefficient of variation, n is the total number of major citrus-producing regions, $x_{i,t}$ represents the green total factor productivity of the citrus industry in region i during period t, and $\overline{x_t}$ is the average value of the green total factor productivity of citrus during period t. If ${\alpha }_{t+1}<{\alpha }_t$, it indicates that the difference in the green development level of the citrus industry is narrowing over time. Conversely, it means that there is a divergence in the green development level of the citrus industry.

β-convergence analysis distinguishes between two distinct types: absolute and conditional convergence. Absolute β-convergence evaluates the catch-up effect in citrus industry green development, where regions with initially lower green total factor productivity (GTFP) exhibit faster growth rates to converge toward more advanced regions. Building upon the methodological framework established by Wang et al. [53], we specify the absolute β-convergence model in Equation (5):

$$ln{x}_{i,t+1}-ln{x}_{i,t}={\alpha }_2+{\beta }_{1}ln{x}_{i,t}+{\epsilon }_{i,t}$$

(5)

The conditional β-convergence analysis explores whether regional citrus industries converge to their own steady-state green development levels while controlling for heterogeneous environmental factors. This method assesses convergence trajectories relative to region-specific equilibrium points, rather than a universal benchmark. Following the econometric specification of Xu et al. [54], the conditional β-convergence model is formalized in Equation (6):

$$ln{x}_{i,t+1}-ln{x}_{i,t}={\alpha }_2+{\beta }_{2}ln{x}_{i,t}+Z_{i,t}+ {\epsilon }_{i,t}$$

(6)

In Equations (5) and (6), x denotes the citrus green total-factor productivity (GTFP), i indexes regions, t indexes years, σ is the constant term, and ε represents the stochastic disturbance term. The vector Z comprises control variables: per-capita disposable income of rural residents (in CNY 10,000) and the regional urbanization rate, both of which proxy the province-level socio-economic development and are sourced from the National Bureau of Statistics of China. The parameter β measures the convergence coefficient. When β₁ is significantly less than 0, it indicates that there is a “catch-up effect” in the level of green development of the citrus industry; conversely, there is no absolute β-convergence. When β₂ is significantly less than 0, it indicates that the level of citrus green development can converge to its own steady-state level with the passage of time; conversely, there is no conditional β-convergence.

4. Indicator Selection and Data Processing

Based on the availability of data, this paper takes the interperiod panel data of eight major citrus-producing provincial administrative regions, namely Chongqing, Hubei, Hunan, Jiangxi, Zhejiang, Fujian, Guangdong, and Guangxi, from 2008 to 2021 as the research sample, and the data were calculated using MaxDEA 9 software. According to geographic regions, Chongqing, Hubei, Hunan, and Jiangxi belong to the main citrus-producing areas in the Yangtze River Region, Zhejiang and Fujian belong to the main citrus-producing areas in the Zhejiang–Fujian Hilly Region, and Guangdong and Guangxi belong to the main citrus-producing areas in the Guangdong–Guangxi Hilly Region [46]. In 2021, the citrus planting area of these eight provincial administrative regions accounted for about 80% of the total citrus planting area in China, which is representative to a certain extent. In this paper, we draw on the expression of Li et al. [39] and take the average of the yearbook data of citrus and orange to represent citrus.

4.1. Selection and Processing of Input Indicators

Drawing on He et al.’s study [6], labor costs, land costs, and material and service costs were selected as input variables.

(1)

Labor input costs were quantified as the total expenditure per hectare, incorporating both implicit and explicit labor components: the opportunity cost of household labor (imputed at prevailing agricultural wage rates), and market-rate expenditures for hired labor;

(2)

Land cost: land cost per hectare is used as a measure, including the rent of transferred land and the discounted rent of self-camping;

(3)

Material and service costs: using material and service costs per mu as a measure, including direct agricultural inputs such as fertilizers, agricultural machinery, and irrigation, as well as indirect inputs such as depreciation and management fees.

In order to eliminate the effects of price fluctuations and ensure the comparability of data across periods, labor costs, land costs, and material and service costs are discounted using the agricultural producer price index with 2007 as the base period.

4.2. Selection and Processing of Output Indicators

Drawing on the studies of Chen et al. and Lin et al. [55,56], carbon sinks in the citrus industry and the output value per hectare of citrus were selected as desired outputs, and surface source pollution and carbon emissions in the citrus industry were selected as undesired outputs.

4.2.1. Desired Outputs

(1) Citrus Carbon Sink. The carbon sink of the citrus industry mainly refers to the total amount of carbon dioxide absorbed by citrus through photosynthesis during the growth process. Referring to the study of Tian et al. [57], the formula for calculating the carbon sink of the citrus industry is shown in Equation (7):

$$C=\sum _{i=1}^n{C}_i\times Q_i\times (1-w)/{\lambda }_i$$

(7)

In Equation (7), C is the total citrus carbon sink; C_i is the carbon sequestration rate per unit of citrus crop; Q_i is the citrus economic yield; w is the water content; and λ_i is the citrus economic coefficient. Due to the lack of detailed information on citrus categorized yield statistics, the economic coefficient and water content could only be estimated based on existing studies. According to Lin [58], the economic coefficient of citrus was taken as 0.14; according to Yu et al. [59] and Dai et al. [60], the water content of citrus was taken as 83%. The carbon uptake rate of citrus was referred to the study of Han et al. [61] and was taken as the common value of the crop, 0.450.

(2) Value of Citrus Production per Hectare. The data on citrus production value per hectare is derived from the National Compendium of Agricultural Product Cost and Benefit Information for 2008–2021. To eliminate the influence of price fluctuations, the 2007-based producer price index for agricultural products was utilized for conversion.

4.2.2. Undesired Outputs

(1) Surface Source Pollution. Surface source pollution is measured by the runoff and leaching loss of nitrogen and phosphorus generated in the production of the citrus industry. Referring to the study of Lin et al. [42], the runoff and leaching losses were analyzed using the discharge coefficient method, and the measurement formula is shown in Equation (8):

$$K=K_0+F\times C$$

(8)

In Equation (8), K is the runoff or leaching loss of total nitrogen and phosphorus (kg/hm²); K₀ is the loss of total nitrogen and phosphorus (kg/hm²) when no fertilizer is applied; F is the total nitrogen and phosphorus applied to citrus orchards (kg/hm²); and C is the runoff loss rate or leaching loss rate of total nitrogen and phosphorus (%). The K₀ and C parameter values were taken from the First National Pollution Source Census-Manual of Fertilizer Loss Coefficients of Pollution Sources from Agricultural Sources.

(2) Citrus Industry Carbon Emissions. The citrus industry inevitably produces carbon emissions in the production process due to the occupation of arable land, irrigation, and other behaviors. With reference to the study of Ding et al. [62], the formula for measuring carbon emissions from the citrus industry is shown in Equation (9):

$$E=\sum T_i\times {\delta }_i$$

(9)

In Equation (9), E is the total amount of carbon emissions from citrus industry; T_i is the amount of carbon emission sources in category i; δ_i is the carbon emission coefficient of carbon emission sources in category i. Combined with the characteristics of citrus production, fertilizer, pesticide, agricultural film, diesel fuel, irrigation, and tilling were selected as carbon emission sources. The emission coefficients and reference sources of each carbon source are shown in

Table 2. Carbon emission coefficients of the main carbon sources in agricultural land use activities.

Carbon Sources	Carbon Emission Factor	Reference Source
Diesel	0.59 kg/kg	IPCC2013
Fertilizers	0.89 kg/kg	Oak Ridge National Laboratory in the United States
Agrochemical	4.93 kg/kg	Oak Ridge National Laboratory in the United States
Agricultural film	5.18 kg/kg	Institute of Agricultural Resources and Ecological Environment, Nanjing Agricultural University
Irrigated	266.48 kg/hm2	Duan et al. [63]
plow	312.60 kg/km2	Li et al. [64]

:

In this study, labor cost, land cost, material and service cost are taken as input indicators of production efficiency in the citrus industry; surface source pollution and carbon emission are selected as undesired output indicators, and citrus carbon sink and citrus output value are selected as desired output indicators. The specific input–output indicator system is shown in

Table 3. Descriptive statistics of input–output indicators of productivity in the citrus industry.

Indicator Name	Maximum	Minimum	Average	Standard Deviation	Unit
labor cost	5.65	0.66	1.49	2	Ten thousand yuan (CNY) per hectare
land cost	0.41	0.06	0.21	0.09	Ten thousand yuan (CNY) per hectare
material and service costs	3.12	0.39	1.28	0.71	Ten thousand yuan (CNY) per hectare
surface pollution	29.27	17.67	22.14	2.99	Kilogram per hectare
carbon footprint	1.68	0.49	1.05	0.33	t per hectare
carbon credits	30.39	4.97	14.83	3.66	t per hectare
value of citrus pro-duction	13.87	1.52	5.37	2.80	Ten thousand yuan (CNY) per hectare

:

5. Empirical Results and Analysis

5.1. Analysis of Technical Efficiency of Citrus Environment Based on Static Perspective

Drawing on a dynamic perspective, this subsection measures and analyses the GTFP of China’s citrus sector using results obtained from the super-efficiency EBM model.

When undesirable outputs are included, environmental pure technical efficiency (GPTE) is estimated under the variable-returns-to-scale (VRS) assumption, following Equation (1). Environmental technical efficiency (GTE) is calculated under the constant-returns-to-scale (CRS) assumption by omitting the constraint $\sum _{j=1,j\ne o}^n{\gamma }_j^t=1$ in Equation (1). Environmental scale efficiency (GSE) is then derived as GSE = GTE ÷ GPTE. If undesirable outputs are excluded, the key parameter ε_b is set to 0. Pure technical efficiency (PTE) is estimated under VRS, whereas conventional technical efficiency (TE) is computed under CRS, again by removing $\sum _{j=1,j\ne o}^n{\gamma }_j^t=1$. Scale efficiency (SE) is obtained as SE = TE ÷ PTE. These calculations yield the data reported in Table 3 and Table 4 below.

5.1.1. Overall Analysis

As shown in Table 4, from 2008 to 2021, the mean value of traditional technical efficiency (TE) of Chinese citrus is 0.992, the mean value of pure technical efficiency (PTE) is 1.067, and the mean value of scale efficiency (SE) is 0.929; after considering the undesired outputs, the mean value of environmental technical efficiency (GTE) is 0.997, the mean value of environmental pure technical efficiency (GPTE) is 1.076, and the mean value of environmental scale efficiency (GSE) is 0.929. Regardless of whether or not undesired outputs are considered, the technical efficiency of citrus in China from 2008 to 2021 is less than 1, which is below the production frontier, indicating that citrus production as a whole is in a relatively ineffective state, and both the traditional technical efficiency and scale efficiency have not reached an effective level, and there is room for improvement. The environmental technical efficiency after considering undesired outputs is slightly higher than the technical efficiency without considering undesired outputs, reflecting the relatively reasonable utilization of resources in Chinese citrus production.

In terms of the time trend, the overall environmental technical efficiency of citrus in China from 2008 to 2021 is relatively stable, located between 0.932 and 1.069. Further observation reveals that the trend of environmental technical efficiency and environmental scale efficiency is more consistent from 2008 to 2018, and the trend of environmental pure technical efficiency is more consistent from 2019 to 2021, indicating that China’s environmental technical efficiency is mainly affected by scale efficiency before 2018, and is mainly affected by pure technical efficiency after that. In particular, 2018 is an important node for the transformation and upgrading of China’s citrus industry, and this change may be related to national policies: since 2017, China’s Ministry of Agriculture (MOA) has organized annual off-the-job business training for no less than one-third of the country’s grassroots agricultural technicians for more than five consecutive days; in the same year, China’s Ministry of Agriculture (MOA), in conjunction with the Development and Reform Commission (DRC), and Forestry Bureau (FB), compiled and issued an Outline of the Plan for Construction of Advantageous Districts of Characteristic Agricultural Products (ADPZs), which included the creation of 12 citrus special-advantage districts, such as Nanfeng honey tangerines and Gannan navel oranges; and the MOA also issued a circular on Guidelines for the Registration of Varieties of Non-Major Crops, proposing to strengthen the registration of citrus varieties and guide the regional agglomeration of the citrus industry and the optimization of social resources.

In terms of sources of efficiency, the comprehensive environmental efficiency of citrus production in China throughout the observation period comes from the combined effect of pure technical efficiency and scale efficiency. The comprehensive efficiency index increases and decreases during the observation period, indicating that the scale of citrus production is unstable, and the utilization efficiency of production means is highly variable. Delineating citrus advantageous areas, upgrading scale production, and increasing citrus research inputs are the keys to improving the comprehensive efficiency of citrus production in China.

Table 4. Environmental technical efficiency and its decomposition of citrus in China from 2008 to 2021.

Period	Without Considering the Undesired Output			Consider the Undesired Output
	TE	PTE	SE	GTE	GPTE	GSE
2008	0.952	1.012	0.943	0.943	1.041	0.906
2009	0.972	1.001	0.972	0.960	0.995	0.962
2010	0.991	1.033	0.959	0.952	1.022	0.932
2011	0.990	1.054	0.937	0.975	1.054	0.923
2012	0.947	1.090	0.863	0.945	1.086	0.867
2013	1.031	1.096	0.947	1.006	1.120	0.912
2014	1.017	1.065	0.951	1.015	1.089	0.936
2015	0.966	1.048	0.921	1.042	1.105	0.946
2016	1.069	1.134	0.943	1.033	1.124	0.921
2017	1.003	1.094	0.912	1.013	1.096	0.926
2018	1.013	1.075	0.940	1.030	1.098	0.943
2019	0.932	1.078	0.859	1.009	1.084	0.933
2020	1.017	1.097	0.923	1.052	1.097	0.962
2021	0.990	1.056	0.937	0.978	1.050	0.935
Average	0.992	1.067	0.929	0.997	1.076	0.929

Table 4. Environmental technical efficiency and its decomposition of citrus in China from 2008 to 2021.

Period	Without Considering the Undesired Output			Consider the Undesired Output
	TE	PTE	SE	GTE	GPTE	GSE
2008	0.952	1.012	0.943	0.943	1.041	0.906
2009	0.972	1.001	0.972	0.960	0.995	0.962
2010	0.991	1.033	0.959	0.952	1.022	0.932
2011	0.990	1.054	0.937	0.975	1.054	0.923
2012	0.947	1.090	0.863	0.945	1.086	0.867
2013	1.031	1.096	0.947	1.006	1.120	0.912
2014	1.017	1.065	0.951	1.015	1.089	0.936
2015	0.966	1.048	0.921	1.042	1.105	0.946
2016	1.069	1.134	0.943	1.033	1.124	0.921
2017	1.003	1.094	0.912	1.013	1.096	0.926
2018	1.013	1.075	0.940	1.030	1.098	0.943
2019	0.932	1.078	0.859	1.009	1.084	0.933
2020	1.017	1.097	0.923	1.052	1.097	0.962
2021	0.990	1.056	0.937	0.978	1.050	0.935
Average	0.992	1.067	0.929	0.997	1.076	0.929

5.1.2. Analysis of Regional Differences

As can be seen from

Table 5. Environmental technical efficiency and decomposition of main citrus-producing regions in China from 2008 to 2021.

Region	Without Considering the Undesired Output			Consider the Undesired Output
	TE	PTE	SE	GTE	GPTE	GSE
Chongqing	0.974	1.063	0.915	0.942	1.049	0.987
Hubei	0.873	1.006	0.869	0.968	1.074	0.904
Hunan	1.129	1.185	0.955	1.074	1.251	0.864
Jiangxi	0.908	1	0.908	0.95	1.024	0.928
Yangtze River Region	0.971	1.064	0.912	0.984	1.1	0.921
Zhejiang	1.068	1.111	0.961	1.035	1.091	0.95
Fujian	0.968	1.012	0.956	0.985	1.035	0.951
Zhejiang–Fujian Hilly Region	1.018	1.062	0.956	1.01	1.063	0.951
Guangdong	1.093	1.155	0.946	1.073	1.087	0.988
Guangxi	0.925	1.003	0.922	0.946	0.996	0.949
Guangdong–Guangxi Hilly Region	1.009	1.079	0.934	1.009	1.042	0.969
National Region	0.992	1.067	0.929	0.997	1.076	0.940

, when undesired outputs are not taken into account, the traditional technical efficiency (TE) of China’s major citrus-producing provincial administrations in 2008–2021 range from 0.873 to 1.129, with Hunan, Guangdong, and Zhejiang ranking among the top three, while Chongqing, Fujian, Guangxi, Jiangxi, and Hubei are below the national average (0.992), and Hubei has the lowest technical efficiency (0.873); after taking into account non-expected outputs, the environmental technical efficiency (GTE) of the major producing provinces are between 0.942 and 1.074. After considering undesired outputs, the environmental technical efficiencies of the main producing provincial administrative regions are located between 0.942 and 1.074, with Hunan (1.074), Guangdong (1.073), and Zhejiang (1.035) in the top three, and Fujian, Hubei, Jiangxi, Guangxi, and Chongqing environmental technical efficiency below the national average (0.997), and Chongqing environmental technical efficiency (0.942) the lowest.

From the source of efficiency, without considering the non-expected output, the traditional pure technical efficiency (PTE) of citrus in China from 2008 to 2021 is located at 1~1.185, and the traditional scale efficiency (SE) is located at 0.869~0.961, and all the main-producing provincial administrations have higher pure technical efficiency than scale efficiency; considering the non-expected output, the environmental pure technical efficiency (GPTE) is located at 0.996~1.251, and the environmental scale efficiency (GSE) is located at 0.996~1.251. Considering undesired outputs, the pure technical efficiency of the environment is located at 0.996~0.251, and the scale efficiency of the environment is located at 0.864~0.988, and the pure technical efficiency of the environment is higher than the scale efficiency of the environment in all the main producing provincial administrative regions. Regardless of whether undesired output is considered or not, pure technical efficiency is the primary source of production efficiency improvement in the main citrus-producing provinces.

The comparative analysis of citrus production efficiency in different producing areas shows that in the traditional efficiency measurement without considering undesired outputs, the three main producing areas show obvious differences. The mean values of traditional comprehensive efficiency in the citrus primary producing areas of the Yangtze River Region, the Zhejiang–Fujian Hilly Region, and the Guangdong–Guangxi Hilly Region were 0.971, 1.018, and 1.009, respectively. From the viewpoint of the efficiency decomposition, the mean value of the pure technical efficiency of each producing area exceeded 1 (1.064 in the Yangtze River Region, 1.062 in the Zhejiang–Fujian Hilly Region, 1.079 in the Guangdong–Guangxi Hilly Region), which indicated that the application of the production technology had already reached a better level; however, the efficiency of the scale was relatively low (0.912 in the Yangtze River Region, 0.956 in the Zhejiang–Fujian Hilly Region, and 0.934 in the Guangdong–Guangxi Hilly Region), reflecting that there is still room for upgrading the scale operation. When considering the non-desired outputs, the environmental efficiency indicators exhibit new characteristics: the average value of environmental comprehensive efficiency in each production area (0.984 in the Yangtze River Region, 1.01 in the Zhejiang–Fujian Hilly Region, 1.009 in the Guangdong–Guangxi Hilly Region) is similar to that of traditional efficiency. However, the pure technical efficiency in the environmental context has further improved (1.100 in the Yangtze River region, 1.063 in the Zhejiang–Fujian Hilly Region, and 1.042 in the Guangdong–Guangxi Hilly Region), which indicates that the application of green production technology plays a significant role in enhancing overall efficiency. The results demonstrate significant improvements in environmental scale efficiency across all regions (0.921 in the Yangtze River Region, 0.951 in the Zhejiang–Fujian Hilly Region, and 0.969 in the Guangdong–Guangxi Hilly Region) compared to traditional scale efficiency measures. These findings indicate that scale operation efficiency under environmental constraints is more favorable.

Notably, the middle and upper reaches of the Yangtze River producing area have the largest increase in environmental efficiency indicators, while the Guangdong–Guangxi Hilly Region has the most obvious improvement in environmental scale efficiency. The rapid improvement of environmental efficiency in the middle and upper reaches of the Yangtze River producing area is mainly reflected in policy-driven technological innovation and the promotion of circular economy models: Hubei Province has implemented the Agricultural Comprehensive Development Yangtze River Green Ecological Corridor Project with loans from the Asian Development Bank, following the concept of green ecology. Through the construction of soil and water conservation projects, it has improved the field microclimate and reduced soil erosion; Zhongxian County, Chongqing has promoted the “biogas slurry + citrus” model, built 90 liquid fertilizer demonstration bases, and absorbed the fecal sewage of 24 farms annually, reducing dependence on chemical fertilizers. In contrast, the Guangdong–Guangxi Hilly Region has achieved significant improvement in environmental scale efficiency through large-scale operation. Data from the Guangdong Provincial Department of Agriculture and Rural Affairs show that the number of citrus planting cooperatives in the province increased by 3.2 times in 2020 compared with 2008, and the average operation scale expanded to 58 mu. This large-scale operation not only improves resource utilization efficiency but also enhances the ability to cope with climate risks. Moreover, the Guangdong–Guangxi Hilly Region has vigorously facilitated the adoption of advanced agricultural technologies and machinery, thereby enhancing agricultural productivity and the efficiency of resource utilization. Guangxi has carried out the machine-friendly transformation of mountain orchard tillage roads and strip terraces, repairing the import and export connections between terrace belts and tillage roads, and widening the operation roads for agricultural machinery on terrace strip fields, which has greatly increased the operable area of agricultural machinery. This has achieved a saving of CNY 1300 in labor input per mu for mango production, and mechanical harvesting of sugarcane saves CNY 300 in cost per mu compared with manual cutting [65].

Figure 1 shows the mean values of GTFP (green total factor productivity) of China’s main citrus-producing areas without considering undesired outputs and with considering undesired outputs from 2008 to 2021.

From a provincial perspective, the technical efficiency of the five primary citrus-producing provinces and autonomous regions—Chongqing, Fujian, Guangxi, Jiangxi, and Hubei—remains below 1, regardless of whether undesired outputs are factored in. This indicates that these regions have yet to attain optimal production states in their citrus industries. In contrast, Hunan, Guangdong, and Zhejiang consistently exhibit leading citrus technical efficiency nationwide. This is attributable to their superior natural and technological endowments. As a thermophilic plant, citrus requires sufficient heat to avoid low-temperature damage during growth, which is crucial for fruit development and ripening. Ample sunlight also enhances photosynthesis, boosting fruit sugar content and quality. Hunan, Guangdong, and Zhejiang, situated in the subtropical monsoon climate zone, benefit from abundant precipitation, heat, and light, thanks to the combined effects of marine water vapor transport and topographic uplift. These climatic advantages are conducive to citrus production. Additionally, the relatively high economic development levels of these provinces provide ample financial resources for technological research and promotion. For instance, the Zhejiang Tangerine Germplasm Research Center collaborates with multiple universities and research institutions to develop new germplasms with desirable traits, construct digital breeding greenhouses, and employ various breeding methods to enhance citrus yield and quality.

From a regional standpoint, the Zhejiang–Fujian Hilly Region and the Guangdong–Guangxi Hilly Region exhibit average comprehensive efficiency of citrus production that surpasses 1. In contrast, the middle and upper reaches of the Yangtze River, with an average efficiency below 1, have yet to attain optimal production levels. The main reasons for this difference are as follows: The Guangdong–Guangxi Hilly Region is recognized as the most suitable area for citrus cultivation in China, featuring favorable temperature and sunshine conditions for citrus growth [66]. In 2022, Guangxi had a citrus planting area of 9.48 million mu and an expected annual output of 18.68 million tons, accounting for 28.37% of the national total and ranking first for eight consecutive years. The Zhejiang–Fujian Hilly Region, being the most economically vibrant among the three regions, enjoys superior scientific and financial support [67]. Conversely, most citrus orchards in the middle and upper reaches of the Yangtze River are situated on steep slopes, primarily consisting of family farms. Technology dissemination is challenging in these areas, and traditional agricultural machinery is often impractical. Moreover, some regions are susceptible to cold waves and pests, leading to fruit drop. In 2016, freezing damage caused a 15% reduction in citrus yield in western Hubei.

Overall, when undesired outputs are not considered in the national main citrus-producing areas, pure technical efficiency is higher than scale efficiency, which is the primary source of improvement. In the comparison of producing areas, the traditional comprehensive efficiency of the middle and upper reaches of the Yangtze River, the Zhemin Hilly Region, and the Liangguang Hilly Region varies significantly. The pure technical efficiency exceeds 1, and the scale efficiency needs to be improved. After considering the undesired outputs, the environmental pure technical efficiency is improved, and the environmental scale efficiency is improved. The middle and upper reaches of the Yangtze River have the largest increase, and the environmental scale efficiency of the Liangguang Hilly Region is significantly improved.

5.2. Analysis of Citrus Green Total Factor Productivity from a Dynamic Perspective

This section adopts a dynamic perspective to estimate and analyze the green total-factor productivity (GTFP) of China’s citrus industry. All results are derived from the GML index calculated via Equation (2); the resulting GML values are then combined with the static estimates and substituted into Equation (3) to obtain the findings reported in Section 5.2.1 and Section 5.2.2.

5.2.1. Overall Analysis

According to

Table 6. Green total factor productivity and its decomposition of citrus in China from 2008 to 2021.

Period	Without Considering the Undesired Output			Consider The Undesired Output
	ML	TC	EC	GML	GTC	GEC
2008–2009	0.939	0.885	1.057	0.917	0.885	1.043
2009–2010	0.980	0.901	1.091	0.956	0.926	1.050
2010–2011	1.123	1.018	1.105	1.081	1.013	1.089
2011–2012	0.965	0.917	1.063	0.942	0.907	1.062
2012–2013	1.168	1.000	1.181	1.106	0.980	1.149
2013–2014	1.160	0.990	1.180	1.120	0.977	1.162
2014–2015	1.038	0.932	1.122	1.060	0.897	1.203
2015–2016	0.977	0.775	1.299	0.984	0.831	1.194
2016–2017	1.040	0.892	1.224	1.037	0.894	1.177
2017–2018	1.082	0.907	1.256	1.097	0.923	1.212
2018–2019	1.032	0.941	1.154	1.057	0.906	1.189
2019–2020	1.021	0.843	1.278	1.048	0.856	1.249
2020–2021	1.011	0.859	1.258	1.038	0.913	1.160
Average	1.041	0.912	1.174	1.034	0.916	1.149

, during the period 2008–2021, China’s citrus green total factor productivity (GML) (1.034), which takes into account non-expected outputs, is slightly lower than the traditional total factor productivity (ML) (1.041), which does not take into account non-expected outputs, suggesting that there is still room for improvement in China’s citrus green production level. The mean value of China’s citrus green total factor productivity is 1.034 (greater than 1), indicating that China’s citrus green total factor productivity is generally on the rise during the period under investigation, and that although there is room for improvement, the overall development of citrus greening is in good shape. From the perspective of the components of green total factor productivity, the technical progress index (GTC) (0.916) is less than 1, and the technical efficiency index (GEC) (1.149) is greater than 1, indicating that: the technological progress in China’s citrus production has decelerated, while technical efficiency has steadily improved. The enhancement of citrus green total factor productivity (GTFP) primarily stems from gains in technical efficiency, with technological progress contributing less significantly. Possible reasons for the diminished role of technological progress include diminishing marginal benefits [68], slow technological updates and application lags, and limited farmer capacity to adopt new technologies [69]. These factors have become key bottlenecks restricting the growth of citrus GTFP.

5.2.2. Analysis of Regional Differences

According to Figure 2, the total factor productivity of major citrus-producing provinces in China from 2008 to 2021 presents the following characteristics:

From the perspective of different provinces, when non-expected output is not considered, the traditional total factor productivity (ML) of only Fujian and Guangxi is greater than 1. The technical progress index (TC) of each province is generally greater than 1, and the technical efficiency index (EC) is generally less than 1. The traditional total factor productivity of each province mainly relies on the technical progress index. After considering non-expected outputs, the green total factor productivity (GML) of Fujian, Guangxi, and Chongqing is greater than 1. The green total factor productivity of each province mainly comes from the green technical progress index (GTC), and the contribution of the green technical efficiency index (GEC) is lower; the green technical efficiency index of only Chongqing Municipality and Fujian Province is greater than 1. In particular, five provinces, Hubei, Hunan, Zhejiang, Jiangxi, and Guangdong, have total factor productivity (TFP) values below 1, regardless of whether non-expected outputs are taken into account, and Fujian and Guangxi continue to lead, which is closely related to their favorable natural conditions, such as precipitation, heat, and light. Considering the non-expected outputs, the GTFP of Chongqing, Jiangxi, Zhejiang, Guangdong, and Guangxi increases, while that of Hubei, Hunan, and Fujian decreases. Among them, five provinces, including Chongqing, maintain efficiency growth after incorporating environmental costs, indicating that their production systems are environmentally adaptive; the efficiency decline in three provinces, including Hubei, reflects the pressure of environmental constraints on their existing production models.

From the perspective of different production areas, the traditional total factor productivity of the Zhejiang–Fujian Hilly Region is higher than that of the Yangtze River Region and the Guangdong–Guangxi Hilly Region, and the green TFP of citrus in Zhejiang–Fujian Hilly Region is greater than 1. The green TFP of citrus in the Yangtze River Region and the Guangdong–Guangxi Hilly Region is less than 1, and the overall trend shows that Zhejiang–Fujian Hilly Region > Guangdong–Guangxi Hilly Region > the Yangtze River Region. Compared with the latter two, the Zhemin Hilly main production area has technical support advantages and financial support advantages. During the study period, the provincial and municipal governments in the Zhemin production area introduced local subsidies for agricultural machinery purchase, digital agriculture, agricultural environmental protection, and agricultural technology research and development and promotion, which significantly promoted the improvement of regional GTFP.

Overall, when undesired outputs are not considered, only Fujian and Guangxi have a traditional total factor productivity (TFP) greater than 1, which mainly relies on the technological progress index. When undesired outputs are considered, Fujian, Guangxi, and Chongqing have a GTFP greater than 1, which mainly comes from the green technological progress index. In terms of production areas, the Zhemin Hilly Region has the highest traditional TFP, the GTFP of the Zhemin Hilly Region has significantly increased, and the green development level of the Liangguang Hilly Region has been continuously improving.

5.3. Convergence Analysis

The above measurements show that there are regional differences in green total factor productivity in China’s citrus industry. In order to deeply analyze its equilibrium, this paper conducts a convergence analysis of green total factor productivity at the national level and in three major citrus-producing regions. Dynamic GTFP measurement offers dual advantages. It not only better accommodates agriculture’s green development complexity, but also more precisely tracking productivity changes under environmental limits. Its inclusion of undesirable outputs further improves real-world applicability. Based on this, this study chooses to analyze the convergence of citrus green total factor productivity, including undesired output, under the dynamic perspective (GMI). Specifically, the σ-convergence results are computed using Equation (4), the absolute β-convergence results using Equation (5), and the conditional β-convergence results using Equation (6). The results of robustness tests are provided in Appendix A.

5.3.1. σ-Test and Result Analysis

Figure 3 depicts the σ-convergence of GTFP in the citrus industry at both the national level and across the three regions. Nationally, the coefficient of variation of GTFP in the citrus industry exhibits fluctuating changes, indicating an overall trend of divergence. At the sub-regional level, the coefficient of variation of GTFP in the Yangtze River Region was consistent with the national trend, but the fluctuation was larger, and the regional differences showed a widening trend; the regional differences of GTFP in the Zhejiang–Fujian Hilly Region were generally decreasing in fluctuation; and the regional differences of GTFP in the Guangdong–Guangxi Hilly Region was increasing in fluctuation. Overall, the regional differences in the level of green development of the citrus industry in the country and the three regions are fluctuating, and do not show significant σ-convergence characteristics.

5.3.2. β-Test and Result Analysis

(1)

Absolute β-convergence

While σ-convergence is characterized by its ease of comprehension and simplicity in calculation, it fails to take into consideration the initial levels of environmental factors as well as the initial factor structures in different regions [70]. To more accurately assess the convergence of GTFP in the citrus industry, this paper further examines the trend of absolute β-convergence in sub-regions, with results presented in

Table 7. Results of absolute β-convergence test for citrus GTFP in China and three major regions, 2008–2021.

Parameters	National Region	The Yangtze River Region	Zhejiang–Fujian Hilly Region	Guangdong–Guangxi Hilly Region
β1	−0.167 *** (0.063)	−0.307 *** (0.100)	−0.065 (0.093)	−0.064 (0.115)
σ1	−0.013 (0.017)	−0.028 (0.021)	0.013 (0.032)	−0.037 (0.037)
White test	0.67	1.20	3.99	2.12
Reset test	0.72	0.77	0.19	0.53
R2	0.431	0.565	0.437	0.5748
F statistic	7.060 ***	9.350 ***	0.490	0.310

Standard errors in parentheses, *** indicates significance at the 1% level.

:

During the study period, the parameter β1 values for the whole country and the Yangtze River Region are significantly negative at the 1% level. This indicates a negative correlation between the GTFP growth rate and the initial level in these regions. It suggests a “catching-up effect,” where lagging provinces narrow the gap with developed ones. This is attributed to the widespread adoption of citrus production technology and improvements in the agricultural technology service system. These factors facilitate the efficient flow of production factors and reduce disparities in green development levels within the production area. Meanwhile, the β1 estimates for the Zhejiang–Fujian Hilly Region and the Guangdong–Guangxi Hilly Region are negative but not significant. This indicates no absolute convergence of GTFP in these areas. This implies that it is difficult to automatically narrow the regional gap by relying only on natural growth mechanisms, such as technology spillovers, and that targeted policies for external intervention are needed.

(2)

Conditional β-Convergence

Table 8. Conditional β-convergence test results for GTFP in China and three major regions, 2008–2021.

Parameters	National Region	The Yangtze River Region	Zhejiang–Fujian Hilly Region	Guangdong–Guangxi Hilly Region
β2	−0.177 *** (0.066)	−0.306 *** (0.104)	−0.305 (0.214)	−0.633 * (0.318)
σ2	0.078 (0.147)	0.940 (0.636)	7.361 (3.486)	0.833 (0.464)
White test	12.01	6.01	13.57	10.43
Reset test	4.37	0.47	0.88	1.51
R2	0.436	0.251	0.370	0.310
F statistic	2.460 *	3.690 **	1.760	1.35

Standard errors in parentheses, *, **, and *** indicate significant at the 10%, 5%, and 1% levels, respectively.

reports the regression results of conditional β-convergence of GTFP for the citrus industry in China and the three major regions from 2008 to 2021. At the national level, its β2 estimates are significantly negative at the 1% statistical level, indicating that there is a tendency for the GTFP of the citrus industry in all provinces of the country to converge to its own steady-state level. At the regional level, the β estimates of the Yangtze River Region was all significantly negative at the 1% statistical level, the β2 estimates of the Guangdong–Guangxi Hilly Region was all significantly negative at the 10% statistical level, and the β2 estimates of the Zhejiang–Fujian Hilly Region was negative but insignificant, indicating that conditional β-convergence of citrus GTFP exists at the national level, in the Yangtze River Region and in the Guangdong–Guangxi Hilly Region, while in the Zhejiang–Fujian Hilly Region, conditional β-convergence of citrus GTFP does not exist.

This suggests that there is a “latecomer advantage” in citrus GTFP growth in the Yangtze River Region and the Guangdong–Guangxi Hilly Region, and that the areas with lower initial GTFP levels within these two regions can catch up with the high GTFP regions at a faster rate through technological learning, policy support, or resource optimization. The difference in citrus GTFP with other regions may persist or even widen in the Zhejiang–Fujian Hilly Region due to the absence of conditional β-convergence. This may stem from the fact that the Zhejiang–Fujian Hilly Region relies more on high value-added varieties, and that land fragmentation and higher labor costs limit the diffusion of technology, resulting in a different GTFP growth path than in other producing regions.

Each region shows different convergence characteristics due to regional heterogeneity in economic, social and resource endowments, and policy measures can be taken to promote GTFP growth in the citrus industry in each production area.

5.3.3. Comprehensive Discussion

Taken together, the convergence tests indicate that regional disparities in China’s citrus green total-factor productivity (GTFP) stem from a complex interplay of factors. From a policy perspective, blanket technology-diffusion initiatives are unlikely to be effective; instead, differentiated interventions are required. For regions exhibiting convergence, priority should be given to strengthening technology-spillover channels, whereas advanced but non-converging regions need support for innovation-led development models. Although the present study safeguards the robustness of its findings, future research with finer-grained farm-level data could open the “black box” of convergence mechanisms.

6. Discussion

6.1. International Comparative Insights

Beyond the domestic empirical analysis, a cross-national perspective is indispensable for deepening our understanding of China’s citrus green development. Drawing on the experiences of leading global citrus producers—Brazil, the United States, and Spain—this section distills transferable lessons to enhance the international relevance and explanatory power of our findings.

Globally, green total factor productivity (GTFP) has emerged as a core metric for assessing agricultural sustainability and resource-use efficiency. Brazil, the world’s largest orange-juice exporter, concentrates its citrus production in São Paulo. Leveraging large-scale plantations and highly mechanized management systems, the country has achieved a synergistic improvement in both technical and scale efficiency [71]. Continuous government support for farm mechanization and export-oriented processing has forged a green, intensive development path in which processing industries are driven by production, and vice versa.

In the United States, California and Florida exemplify the adoption of digital agriculture and precision irrigation. Sensor networks, remote sensing, and drones enable end-to-end regulation of water and nutrients and intelligent forecasting of pest and disease outbreaks, substantially reducing non-point-source pollution. For instance, California citrus orchards have used drone-based precision spraying to cut labor costs by more than 35% [72]. The U.S. also enforces stringent pesticide control; Mexico complements this by standardizing production processes to reduce pesticide dependence and improve export safety [73].

Spain, a major EU citrus exporter, relies on a dual system of geographical indications and green certification to upgrade its industry. Institutionalized environmental policies and carbon-footprint oversight, integrated with the EU Emissions Trading System, incentivize growers to participate in ecological-compensation schemes, thereby enhancing the market conversion of ecological value.

In comparative terms, China’s citrus sector exhibits advantages in pure technical efficiency but still lags in scale efficiency, fine-grained management, and environmental governance. Drawing on these international experiences, China should prioritize the mechanization and road-infrastructure upgrading of hilly production areas, accelerate the adoption of green technologies in post-harvest handling and standardization, refine green certification and carbon-sink mechanisms, and expedite the transition from resource-input-driven to ecological-value-driven growth.

6.2. Main Findings and Limitations

Unlike previous studies on citrus production efficiency, this paper employs both the super-efficiency EBM model and the GML index to evaluate citrus performance from static and dynamic perspectives, and it contrasts conventional efficiency (which disregards undesirable outputs) with green efficiency (which explicitly accounts for them). Undesirable outputs incorporate not only citrus-related carbon emissions but also non-point-source pollution, while desirable outputs extend beyond citrus output value to include ecological benefits such as carbon sequestration. Finally, σ- and β-convergence tests are conducted for China as a whole and for each major producing region. Through this multi-dimensional measurement framework, the study systematically reveals the spatiotemporal characteristics of China’s citrus green total factor productivity and explores the evolving regional disparities via convergence analysis.

However, data availability and methodological choices impose several limitations that point to avenues for future refinement. First, convergence metrics can be further refined. Relying solely on the coefficient of variation for σ-convergence may obscure changes in absolute dispersion; presenting both the coefficient of variation and the standard deviation in tandem is recommended to capture simultaneous dynamics in absolute and relative gaps. β-convergence has thus far been examined only through unconditional tests at the national and three major regions; future work should incorporate a spatial-weight matrix to implement spatial β-convergence models and test for spill-over effects between neighboring producing areas. Second, the analytical framework for explaining the mechanisms and for performance evaluation needs to be expanded. Although the pronounced “catch-up effect” in the upper–middle Yangtze region is robust, the heterogeneity in convergence speed remains unexplained with respect to policy intensity, resource endowment structures, and market competition patterns. Interaction terms or stratified regressions should be employed to disentangle these drivers. Moreover, the current efficiency measurement omits ecosystem-service values; subsequent studies could integrate carbon sequestration, water conservation, and other ecological outputs into the input–output table, thereby constructing an integrated economic–ecological frontier that aligns evaluation results more closely with sustainable-development objectives.

Hence, future research should pursue two overarching lines of inquiry. First, refine convergence measurement by supplementing the coefficient of variation with standard-deviation metrics and spatial β-convergence tests to capture absolute gaps, neighborhood spillovers, and regional heterogeneity. Second, deepen the analysis of underlying mechanisms and broaden the evaluation framework by quantifying the effects of policy, resource, and market factors while incorporating ecosystem services; together, these steps will support the construction of an integrated economic–ecological efficiency frontier and foster the high-quality and sustainable co-development of the citrus sector.

7. Conclusions and Suggestion

7.1. Research Conclusions

Based on the panel data of eight major citrus-producing provincial administrative regions in China from 2008 to 2021, this paper adopts the super-efficiency EBM model and the GML index to analyze the green total factor productivity of the citrus industry in the major citrus-producing provincial administrative regions in a static and dynamic way, and based on this, it applies the σ-convergence and β-convergence methodology to analyze the convergence of green total factor productivity in the citrus industry, and arrives at the following research conclusions:

First, from the static perspective, the production technology efficiency of the vast majority of major citrus-producing provinces in China have not reached the optimal state, and pure technology efficiency is the primary source of improving production technology efficiency.

Second, from a dynamic perspective, the green total factor productivity of China’s main citrus-producing regions during the study period had large regional differences, with Fujian, Guangxi, and Chongqing having a green total factor productivity greater than 1, and Hubei, Guangdong, Hunan, Zhejiang, and Jiangxi having a green total factor productivity less than 1. In terms of each producing region, the trend of Zhejiang–Fujian Hilly Region > Guangdong–Guangxi Hilly Region > The Yangtze River Region was presented, and the interregional green development was poorly coordinated; the technological efficiency was the primary source of the total factor productivity improvement of Chinese citrus, and the contribution of technological progress to China’s citrus total factor productivity is low.

Third, in terms of convergence, the GTFP of citrus at the national level and in the three major producing regions does not show σ-convergence. β-convergence: both absolute β-convergence and conditional β-convergence exist at the national level and in the Yangtze River Region, while there is no β-convergence in the Zhejiang–Fujian Hilly Region, and only conditional β-convergence exists in the Guangdong–Guangxi Hilly Region, but no absolute β-convergence exists. The existence of absolute β-convergence at the national level and in the Yangtze River Region indicates that there is a “catching-up effect” between the provinces lagging behind in citrus green TFP and the developed provinces in the Yangtze River Region, and that GTFP in the citrus industry in the Yangtze River Region will converge to the steady-state equilibrium value. Conditional β-convergence exists at the national level, in the Yangtze River Region and in the Guangdong–Guangxi Hilly Region, indicating that the GTFP of the citrus industry in the provinces within these regions tends to converge to their respective steady state levels.

7.2. Suggestion

Empirical findings indicate that raising China’s citrus green total factor productivity (GTFP) requires a multi-pronged, high-quality green transition. Priority should be given to: (1) accelerating green-technology innovation, (2) formulating differentiated yet coordinated regional strategies, (3) institutionalizing long-term safeguards for green development, and (4) deepening international cooperation to enhance global competitiveness. Specific policy recommendations are as follows.

7.2.1. Strengthen Green-Technology Innovation to Foster “Efficiency-Plus-Technology” Synergy

Results indicate that pure technical efficiency is currently the chief engine of citrus GTFP growth, yet the contribution of technological advances remains modest. To address the low penetration of green technologies, integrated technology solutions should be prioritized. In provinces such as Hubei and Hunan—characterized by dense water networks and production systems with substantial environmental externalities—vigorously promote a “green package” combining organic-fertilizer substitution and eco-friendly pest-and-disease control, establishing a three-in-one model of “fertilizer reduction–pollution prevention–quality enhancement” [74]. In technology-dependent regions such as the Zhejiang–Fujian hills, accelerate the adoption of “smart orchard management systems” that leverage IoT for precision delivery of water, fertilizer, and pesticides. Create a nationwide green-technology supply-and-demand platform that brings research institutes and firms together to break through key bottlenecks, e.g., mechanization of hillside orchards and post-harvest commercial handling, and forge a full-cycle mechanism of “technology development → pilot validation → demonstration and extension” [75].

7.2.2. Design Differentiated Regional Pathways and Promote Interregional Coordination

In light of pronounced regional disparities in green-development performance, differentiated pathways coupled with coordinated interregional efforts are essential. The upper–middle Yangtze region—marked by rugged terrain and weak infrastructure—should prioritize accelerated investment in irrigation works and hillside-orchard roads to improve production conditions and disaster resilience [76]. In Hubei, Chongqing, and similar localities, terraced orchards retrofitted with sprinkler irrigation can raise land-use intensity and mechanization rates. Hilly provinces such as Zhejiang and Fujian should promote moderate-scale production to offset efficiency losses caused by fragmented management; for instance, the upper–middle Yangtze citrus belt and the Yunnan early-ripening mandarin base have already lowered unit production costs through intensified planting [77]. Provinces with lagging GTFP growth should be targeted with a “technical assistance + fiscal preference” package, leveraging demonstration bases in the east to catalyze technology upgrading in central and western orchards.

7.2.3. Institutionalize Long-Term Green-Development Safeguards

Building on existing agricultural policies, complement them with a trinity of “assessment–compensation–finance” instruments. First, incorporate GTFP growth into provincial rural-revitalization performance indicators with differentiated weights. Second, refine ecological-compensation schemes by emulating Zhejiang’s “eco-compensation + carbon-sink trading” model and establishing trans-provincial watershed “water-quality betting” agreements that grant horizontal fiscal rewards to jurisdictions meeting water-quality targets. Third, innovate green-finance instruments—such as citrus-carbon-sink collateralized loans, green bonds, etc.—and offer concessional interest rates to operators adopting green technologies. Deploy a dynamic monitoring system using satellite remote sensing to track orchard management in real time, ensuring that policy support is precisely targeted.

7.2.4. Deepen International Cooperation to Enhance Global Competitiveness

Benchmarking leading citrus nations such as Brazil and the United States, China should close key value-chain gaps. In processing, adopt Spain’s “geographical indication + carbon label” dual-certification system and build a national citrus carbon-footprint database to provide green passports for export products. Emulate U.S. precision-agriculture technologies by accelerating the development of UAV crop protection and AI pest-and-disease recognition, aiming to double smart-equipment coverage in major producing regions within five years. Leverage the RCEP framework to deepen cross-border capacity cooperation—establish overseas citrus parks in Southeast Asia that export China’s green cultivation technologies and standards. Finally, create a global green-agriculture technology intelligence network that periodically releases frontier-technology briefings, guiding domestic firms to secure the technological high ground.