Introduction

The ecological and environmental issues caused by the continuous increase in the carbon emissions accompanying industrialization and urbanization have become increasingly prominent, attracting widespread attention from governments and scholars in various countries. Some countries, such as the UK and Japan, have put into practice actions and plans to achieve energy conservation and emission reduction. Following the reform and opening up, China has become the world’s largest carbon emitter and has long regarded energy conservation and emission reduction as a national development strategy. In agriculture, China has made a significant achievement by feeding 20% of the world’s population with 7% of the world’s arable land, reaching a sufficiently high level to satisfy the rising national demand for grain (Jin et al. 2019). Nevertheless, this achievement has resulted in prominent ecological environment issues, such as the degradation of cultivated land and organic material, a decrease in basic soil fertility, and the excessive consumption of agricultural chemical material. Here, agricultural carbon emissions refer to carbon emissions that are directly or indirectly caused by the consumption of fertilizers, pesticides, and energy in the process of agricultural production and land plowing. Unreasonable agricultural production activities make agriculture the second main source of emissions in China. According to the estimated data of the Food and Agriculture Organization of the United Nations, the total carbon emissions in agricultural fields account for approximately one-third of the global total. Agricultural carbon emissions caused by agricultural production activities account for 17% of the national total carbon emissions in China, which is second only to the carbon emissions caused by energy consumption activities. Furthermore, methane and nitrogen dioxide emitted by agriculture production activities account for 50% and 92% of the national total, respectively. It must be recognized that the traditional agricultural production model characterized by high inputs, high consumption, and low efficiency has become unsustainable. To respond to the severe carbon emissions associated with agriculture, the Chinese government has applied the strategy of zero growth in the usage of chemical fertilizers and pesticides, advocating the use of “grain storage in the land” and “grain storage in technology” to address practical problems. Moreover, agricultural subsidies have been provided to enhance agricultural production capacity and reduce agricultural production risk and to encourage farmers to adopt practices to protect agricultural resources. To address these issues, this study applies agricultural carbon emissions as an indicator that affects the agricultural production environment for the study of agricultural eco-efficiency. The results may provide important practical guidance to achieve the dual goals of agricultural carbon emission reduction and the transformation of green agriculture.

The agricultural production zone in this study is perceived as a natural agricultural space occupied by cultivated land with advantageous agricultural conditions, high resource endowment, and high productivity. In 2011, China officially issued the national policy of major function zoning, a national territorial development plan intended to achieve a sustainable geographic and landscape pattern (Fan and Li 2009). To manage the national territorial space, this plan creates a new organizational unit of regional functions (Wang and Fan 2020) involving three spatial organizations: the urbanization zone, the agricultural development zone, and the ecological security zone (Fan et al. 2012). For the agricultural development zone, the plan creates a new agricultural spatial organization with seven districts and twenty-three zones to ensure the safety of agricultural products in China. Specifically, as agricultural production zones are primarily oriented towards agriculture, their principal functions concern the effective supply of agricultural products and the comprehensive agricultural production capacity. With the gradual strengthening of resource and environmental constraints, the national agricultural economic development goals involve improvements not only in grain production and the quantity of agricultural products but also in the quality of agricultural economic growth to achieve a balance between the agricultural economy and resource supply. Accordingly, facing the severe constraints of the agricultural environment, promoting agricultural eco-efficiency, and achieving green agriculture are increasingly seen as the research frontier in geography and economics. In this context, how to coordinate the relationship between agricultural production and ecological civilization construction needs to be studied in depth. A comprehensive study focusing on how influencing factors affect agricultural eco-efficiency over time would thus have great practical significance and could provide references for the agricultural policy makers.

Jilin agricultural production zone, a crucial component of the Northeast agricultural production zone that is recognized worldwide for its advantages in commodity grain cultivation, has strategic significance for national grain security in China. Thus, it is adopted as our empirical study area. Facing the situation of resource restraints and increasing grain demand, what are the spatiotemporal characteristics of agricultural eco-efficiency in the Jilin agricultural production zone during the period 2005–2017? How do influencing factors accelerate or hinder the enhancement of agricultural eco-efficiency in this agricultural production zone? What corresponding policy implications can be obtained from the empirical results to support policy makers? The findings regarding these key scientific issues can provide a theoretical reference with practical significance for accomplishing the country’s carbon emission reduction targets ahead of schedule, optimizing the allocation of agricultural resources and promoting the green agricultural transition.

The remainder of this study is organized in five sections. The “Literature review on agricultural eco-efficiency” section reviews the relevant literature and the theoretical analysis on agricultural eco-efficiency. The “Data, methodology, and variable selection” section introduces the study area, data sources, methodology, and variable selection for empirical analysis. The “Empirical results” section provides an assessment of agricultural eco-efficiency, illustrates the spatial dynamics characteristic of agricultural eco-efficiency, and interprets the driving mechanism. The “Conclusion and policy suggestions” section describes the relevant conclusions and proposes corresponding policy suggestions for policy makers.

Literature review on agricultural eco-efficiency

Eco-efficiency emerged in the 1990s, when Schaltegger and Sturm (1990) first proposed the concept as a quantitative tool of environmental management. In 1998, the Organization for Economic Co-operation and Development mentioned the concept of eco-efficiency with the purpose of addressing the relationship between environmental impacts and agricultural production (Camarero et al. 2013). Eco-efficiency conventionally refers to a process that seeks to maximize economic effectiveness while minimizing environmental impacts (Sinkin et al. 2008; Burnett and Hansen 2008). Over time, eco-efficiency has been conceptualized and considered a topic of interest in an increasing range of fields (Reith and Guidry 2003). The international and national research content of eco-efficiency is multi-dimensional and diversified (Zhang et al. 2008), covering the aspects of eco-efficiency evaluation (Huang et al. 2018; Czyżewski et al. 2019; Baum and Bieńkowski 2020), the temporal evolution characteristics and spatial differentiation of eco-efficiency(Liu et al. 2020a; Chen et al. 2017), the influencing factors leading to changes in eco-efficiency(Moutinho et al. 2020), the evolution pattern of eco-efficiency, and strategies supporting the improvement in eco-efficiency. As the understanding of eco-efficiency grows, the number of related studies focusing on specific industries is gradually increasing. International and Chinese scholars have increasingly applied the lessons learned in the economic sector to other industrial sectors from different angles (Lio and Hu 2009). For the eco-efficiency evaluation of a specific industry, studies have gradually come to concentrate on specific kinds of eco-efficiency, considering eco-efficiency at the regional level (Zhou et al. 2020), in the tourism sector (Gössling et al. 2005; Liu et al. 2017; Peng et al. 2017), in various economic sectors (Xing et al. 2018), and in urban areas (Yin et al. 2014; Ren et al. 2019).

Among the in-depth studies on eco-efficiency, many have centered on agricultural production and grain security in the agricultural field. For instance, Picazo-Tadeo et al. (2011) assessed farming eco-efficiency applying data envelopment analysis (DEA) techniques. Gómez-Limón et al. (2012) evaluated the farm-level eco-efficiency among Andalusian olive farmers. Vlontzos et al. (2014) evaluated the agricultural energy and environmental efficiency of EU countries using the DEA approach. Todorovic et al. (2016) conducted the eco-efficiency assessment of agricultural water systems at the meso-level by using the life cycle system-based approach. Saravia-Matus et al. (2019) measured the relationship between greenhouse gas efficiency and agricultural production in the agricultural sector. Deng and Gibson (2019) estimated the agricultural eco-efficiency of Shandong in 1990–2010 based on stochastic frontier analysis. By applying the nonseparable hybrid DEA model considering undesirable outputs, Han and Zhang (2020a) evaluated environmental efficiency and the total factor productivity of cultivated land use. Moreover, scholars of agricultural economics have made considerable efforts to explore the influencing factors that shape the spatiotemporal distribution characteristics of agricultural eco-efficiency. Gkiza and Nastis (2017) empirically verified the effect of human health on agricultural production efficiency. Czyżewski et al. (2020) examined the effect of the European Union Common Agricultural Policy on environmental sustainable value, confirming that the higher investment support and capital–labor ratio contributed to eco-efficiency. Coluccia et al. (2020) assessed the eco-efficiency of the Italian agricultural sector and demonstrated that the Common Agricultural Policy weakened the specific environmental externalities via environmentally friendly land use management. These studies apply a wide range of methods to comprehensively measure the eco-efficiency level, spatiotemporal characteristics, and influencing factors of agriculture from different perspectives (Ma et al. 2018a; Ma et al. 2018b). Zeng et al. (2020) analyzed the relationship between crop diversity and agricultural eco-efficiency and found that crop diversity has a positive effect on agricultural eco-efficiency overall in China and important regional differences were apparent in 34 provinces. Liu et al. (2020b) estimated the agricultural eco-efficiency of Chinese provinces over the period 1978–2019 and found that the current agricultural eco-efficiency had a declining trend and efficiency had remarkable staging features including free development, reform promotion, market regulation, and policy incentives. Liao et al. (2021) evaluated the spatial differentiation characteristics and driving factors of agricultural eco-efficiency and found that the average agricultural eco-efficiency level can still be improved and the agricultural eco-efficiency development of each province is not balanced.

Among the models used for assessment, the ecological footprint method (He et al. 2016; Yang and Yang 2019), ratio method (Schmidheiny and Timberlake 1992), SFA, and DEA are widely applied for the analysis of eco-efficiency, incorporating a multitude of input and output indicators. The calculation of the ratio method is simple and easy to understand. The major advantage of the DEA model over the SFA is that it can effectively eliminate the effect of random errors, and the function form need not be set in advance. Hence, various scholars have increasingly applied the DEA model and improved DEA model to measure agricultural eco-efficiency. Summarizing the existing research methods on efficiency evaluation, DEA includes CCR-DEA, BBC-DEA, SBM-DEA, super DEA, and super SBM-DEA. In recent years, more leading-edge methods with higher measurement accuracy, like the minimum distance to strong efficient frontier with undesirable outputs (MinDS-U) model (Aparicio et al. 2007; Han and Zhang 2020b) and four-step method for the carbon footprint (CF) + DEA approach (Angulo-Meza et al. 2019), have begun to be promoted and employed to assess the eco-efficiency level in the academic field. Additionally, various methods have been applied to verify the influencing factors of agricultural eco-efficiency. Among these methods, the panel data regression method and Tobit regression method are regarded as conventional instruments for analysis to identify the influencing factors, which provide significant reference by revealing the temporal variation trends of agricultural eco-efficiency.

Previous research enriches the understanding of agricultural eco-efficiency, both theoretically and practically. However, empirical studies on agricultural eco-efficiency that focus on the main production area of agricultural products remain scarce, and spatially focused studies on agricultural production zones from the perspective of major function-oriented zones are lacking. Accordingly, this study offers a potential contribution to the existing literature in two aspects. Existing studies have concentrated on the spatial dimension of agricultural eco-efficiency at the national level, provincial level, and city level but not the county level. Liu et al. (2020b) found that agricultural eco-efficiencies in Northeast, East, and South China were higher than that of the national average level, North and Central China fit the national average level, while Southwest and Northwest China were lower than the national average level. Maia et al. (2016) assessed the agricultural eco-efficiency in the Monte Novo irrigation perimeter, which lacks analysis of spatial heterogeneity of agricultural eco-efficiency. Using the data available, this study aims to examine the spatiotemporal characteristics of agricultural eco-efficiency at the county level. In addition, in contrast with the traditional consideration of agricultural eco-efficiency that ignores resource and environmental factors, the improved assessment of agricultural eco-efficiency in this study accounts for the negative impact of resource constraints to accurately reflect the performance of agricultural economic growth. The findings regarding these key scientific issues can provide a theoretical reference with practical significance for accomplishing the country’s carbon emission reduction targets ahead of schedule, optimizing the allocation of agricultural resources and promoting the green agricultural transition. As such, this study not only reveals the characteristics of agricultural eco-efficiency over time and space but also estimates the potential influencing factors to propose suggestions for policy makers and agricultural managers.

Analytical framework

In the interactions among economic, social, and environmental systems, different agricultural production conditions and human development factors are intertwined, complicating the change process of the spatiotemporal pattern of agriculture eco-efficiency. On the one hand, the ratio and scale of agricultural input and output directly affect agricultural eco-efficiency. Agricultural practitioners, the actual implementers of modern agricultural production, determine the amount and structure of input factors, such as land use structure, planting structure, farmland management scale, farming methods, and level of production. Therefore, with the conversion between input and output, changes in the input-output structure can directly cause changes in agricultural eco-efficiency by affecting the allocation and utilization of resources for agricultural production. On the other hand, agricultural eco-efficiency is also indirectly affected by changes in external socio-economic conditions. For example, the economic development level, the transfer of rural laborers to cities, agricultural policies, and agricultural market conditions also significantly affect agricultural eco-efficiency at the macro level. This study intends to explore the complex relationship between agricultural production and agricultural eco-efficiency by considering both its direct and indirect influences. To support this in-depth understanding, an analytical framework illustrating the interactions between agricultural production and agricultural eco-efficiency is proposed in Fig. 1.

Fig. 1
figure 1

An analytical framework illustrating the interactions between agricultural production and agricultural eco-efficiency.

Full size image

Data, methodology and variable selection

Study area and data sources

This study adopts the agricultural production zone of Jilin Province (JAPZ) as the empirical study area. According to the Major Function Oriented Zone Plan of China and Jilin Province, 28 counties in Jilin form this major grain production zone, covering 102,598.59 km2 and occupying 53.52% of the provincial territory. With a population of 18.713 million, which represents 70.3% of the total population in the province, the JAPZ generated 808.319 billion yuan (57.48%) of the province’s GDP in 2018. Notably, given the lack of statistical data, the Shuangyang District and Jiutai District of Changchun City and Taobei District of Baicheng City are not considered in our empirical study. This study evaluates agricultural eco-efficiency at the county level and estimates its influencing factors by employing a panel dataset composed of 26 counties during the 2005–2017 period. Fig. 2 presents a map of the empirical study area. Original socioeconomic data are compiled from the Statistical Yearbooks of Jilin Province, Changchun City, Jilin City, Siping City, Songyuan City, Baicheng City, Liaoyuan City, and Tonghua City for 2006–2018.

Fig. 2
figure 2

Map of the empirical study area.

Full size image

Variable selection

Dependent variable: agricultural eco-efficiency

The underlying principle of agricultural eco-efficiency is to create agricultural economic value with less agricultural input while continuously reducing the effects on the ecology and natural environment. Thus, the calculation of agricultural eco-efficiency integrates three dimensions: agricultural production inputs, desirable agricultural outputs, and undesirable agricultural outputs. Specifically, a conventionally used strategy is established to measure the production factor inputs using the indicators of the labor force, machinery, energy, irrigation, chemical fertilizers, pesticide, and plastic membrane. The desirable outputs comprise two types of agricultural output in the agricultural production zone, namely, agricultural output and grain production capacity. In combination with the actual situation of regional agricultural development, rapid urbanization development has caused the decline of soil quality, which leads the national government and Jilin Province paid increasing attention to the protection of black soil resources and quality. Therefore, we emphasize the agricultural input intensity and agricultural output intensity when selecting input and output indicators. Table 1 displays the agricultural eco-efficiency evaluation index system. The land input is reflected in the human and material inputs per unit of cultivated land. That is, the agricultural eco-efficiency evaluation index system contains the intensity of agricultural production input per unit of cultivated land, the intensity of desirable output per unit of cultivated land, and the intensity of carbon emission pollution per unit of cultivated land to reflect the actual situation of the region. Specifically, this paper selects the total agricultural labor per unit of cultivated land as the labor force indicator, the total agricultural machinery power per unit of cultivated land as the agriculture machinery indicator, the amount of agricultural diesel consumption per unit of cultivated land as the energy consumption indicator, effective irrigation area per unit of cultivated land as the irrigation indicator, the amount of agricultural chemical fertilizers per unit of cultivated land as the chemical fertilizer indicator, the amount of agricultural pesticide usage per unit of cultivated land as the pesticide indicator, and the amount of agricultural plastic membrane per unit of cultivated land as the plastic membrane indicator. Additionally, this paper selects the total gross output value of agriculture per unit of cultivated land as the agricultural output indicator and the total grain output per unit of cultivated land as the grain production capacity indicator. For the undesirable outputs, this study employs the amount of agricultural carbon emissions per unit of cultivated land as the proxy measure. Referring to previous scholarly work (Tian et al. 2014; West and Marland 2002), this study selects agricultural chemical fertilizers, pesticides, plastic sheeting, diesel oil, irrigation, and tillage as the carbon sources of agricultural production activities. Their emission coefficients are 0.8956 (kg/kg), 4.9341 (kg/kg), 5.18 (kg/kg), 0.5927(kg/kg), 266.48(kg/hm2), and 312.6 (kg/km2), respectively. We multiply the emission coefficients by the usage amount or acreage to calculate the total agricultural carbon emissions (Dubey and Lal 2009; Huang et al. 2019). Table 2 presents a statistical description of the indexes used for assessing agricultural eco-efficiency.

Table 1 Agricultural eco-efficiency evaluation index system.
Full size table
Table 2 Descriptive statistics of the agricultural eco-efficiency assessment indexes.
Full size table

Independent variable selection

Multiple-crop index (MI)

The multiple-crop index refers to the frequency of planting crops per unit of cultivated land. This study uses the proportion of the crop sown area to the cultivated land area as the proxy variable of the multiple-crop index.

Scale of family farmland management (FFMS)

Agricultural eco-efficiency also affects the environment through the expansion of the family farmland management scale. As the scale of production and operation units reaches an appropriate level, the allocation of production factors achieves the best operating efficiency, which leads to changes in production, life, ecology, and services. This study thus applies the ratio of the sown area to the number of rural households to characterize the scale of family farmland management.

Agricultural technology extension level (ATEL)

Agricultural technology extension can indirectly guide agricultural production in a more environmentally friendly direction. Due to the unavailability of related data in the statistical yearbooks, the number of agricultural professional and technical personnel is applied as the proxy variable of the agricultural technology extension level.

Agricultural economic development level (AEDL)

The agricultural economic development level may have a close relationship with agricultural environmental quality according to the theory of the environmental Kuznets curve (Grossman and Krueger 1995; Ali et al. 2019). This study, therefore, applies the per capita agricultural output value to represent the agricultural economic development level.

Agricultural industrial structure (AIS)

Compared with the non-planting industries, planting industries may have a greater impact on the agricultural ecological environment through high labor input and labor intensity. Thus, this study applies the proportion of the planting industry in the primary industry as the proxy variable of the agricultural industrial structure.

Urbanization level (UL)

The transfer of the non-agricultural population can lead to changes in agricultural eco-efficiency. Thus, the urbanization level, which here refers to the urban population percentage of the total population, is employed as the proxy variable of the urbanization level.

Level of rural resident income (RRIL)

High resident income generates an income effect and a substitution effect on agricultural eco-efficiency and then prompts rural residents to increase the input of production factors, inevitably resulting in agricultural emissions. However, an improvement in rural resident income level may enable farmers to afford high-quality production factors, thereby decreasing agricultural pollution. The per capita net income of rural residents is regarded as the proxy variable of rural resident income level.

Methodology specification

Measuring agricultural eco-efficiency: super SBM-DEA

The data envelopment analysis (DEA) model proposed in 1978 is an extensively used linear programming technique that can effectively evaluate the relative efficiency of decision-making units (DMUs). The conventionally used DEA model contains the CCR-DEA model and the BCC-DEA model. The former supposes that the returns to scale are constant (Charnes et al. 1978), while the latter supposes that the returns to scale are variable (Banker et al. 1984). Both of these conventional DEA models are radial and oriented and thus overestimate efficiency. As such, Tone (2001) proposed a non-radial and non-oriented slack-based model (SBM) able to account for slackness, which can directly overcome the input and output slacks in the measurement. However, it is possible for multiple DMUs to have valid effective status denoted by 100% at the same time, which makes it difficult to rank and compare the efficiency of DMUs (Färe et al. 1989). To address this issue, Tone (2002) extended the model, proposing the super SBM-DEA model to innovatively solve these disadvantages of the traditional SBM-DEA model. The super SBM method can provide a clear ranking based on the effective agricultural eco-efficiency scores (Li et al. 2013). Thus, this study applies the counties as the DMUs of the agricultural development frontier, using the data for 26 counties in 2003–2016. The formula of the super SBM-DEA model is shown as follows:

$$ {\displaystyle \begin{array}{c}\min \rho =\frac{\frac{1}{m}\sum \limits_{i=1}^m\frac{s_i^{-}}{x_{ik}}}{\frac{1}{s_1+{s}_2}\left(\sum \limits_{r=1}^{s_1}\frac{s_r^g}{y_{rk}^g}+\sum \limits_{q=1}^{s_2}\frac{s_r^b}{y_{qk}^b}\right)}\\ {}s.t.\left\{\begin{array}{c}\sum \limits_{j=1,\ne k}^m{\lambda}_j{x}_{ij}-{s}_i^{-}\le {x}_{ik}\\ {}\sum \limits_{j=1,\ne k}^n{\lambda}_j{y}_{rj}^g+{s}_r^g\ge {y}_{rk}^g\\ {}\sum \limits_{j=1,\ne k}^n{\lambda}_j{y}_{qj}^b-{s}_r^b\le {y}_{qk}^b\\ {}{s}_i^{-}\ge {x}_k,{s}_r^g\le {y}_k^g,{s}_r^b\ge {y}_k^b,{\lambda}_j\ge 0\end{array}\right.\end{array}} $$
(1)

where objective function ρ is the agricultural eco-efficiency and its variation range can be more than 1; if ρ>1, and the sx m, sy n, and sb i=0, the DMU is effective; if 0≤ρ<1, the DMU is ineffective, and the input and output should be improved. λj denotes the coefficient; m is the number of input indicators; s1 is the number of desirable output indicators; s2 is the undesirable output indicators; s- i, gg r, and sb r are the slack variables; and x ik, yg rk, and yb qk denote the ith input, the rth desirable output, and the qth undesirable output value of county k. It is noted that the super SBM-DEA model assumes that there are constant returns to scale.

Verifying the influencing factors: panel data regression model

Panel data, also called time series and cross-sectional data or pooled data, are two-dimensional data obtained over time and across space (Zhou et al. 2018). The panel data regression method can simultaneously reflect the changing pattern and characteristics of variables across the two dimensions of time and space, control individual heterogeneity and endogeneity problems, and improve the effectiveness of parameter estimation. Therefore, this method is widely used for modeling economic problems. The model is defined by the following formula:

$$ {Y}_{\mathrm{it}}={\beta}_0+{\beta}_1{X}_{it}+{\beta}_2{X}_{it}+{\beta}_3{X}_{it}+\kern0.5em +{\beta}_n{X}_{it}+{\beta}_{n+1}{X}_{it}+{\varepsilon}_{it} $$
(2)

where Yit is the dependent variable; Xit is the independent variable; β0 denotes the constant; β1, β2, … βn, βn+1 represent the regression parameters; ε is the random error; i represents the county; and t denotes the time.

To eliminate the heteroscedasticity of variables, we take the natural logarithm of the original data for further conducting the panel data regression model:

$$ lnAE{E}_{\mathrm{it}}={\beta}_0+{\beta}_1 lnM{I}_{it}+{\beta}_2 lnFFM{S}_{it}+\kern0.5em \dots +{\beta}_6 lnU{L}_{it}+{\beta}_7 lnRRI{L}_{it}+{\varepsilon}_{it} $$
(3)

where the variables MI, FFMS, …, UL, RRIL have the same implications as in the “Independent variable selection” section and in Formula (2). Table 3 presents the descriptive statistics of the dependent and independent variables used in this empirical study.

Table 3 Descriptive statistics of dependent and independent variables.
Full size table

Empirical results

According to the abovementioned model specification, this study evaluates the agricultural eco-efficiency using the super SBM-DEA approach that incorporates agricultural carbon emissions and then investigates its influencing factors using the panel data regression method.

The temporal variation characteristics of agricultural eco-efficiency

Based on Formula (1) and the variables selected in the “Dependent variable: agricultural eco-efficiency” section, which include the indicators for seven inputs and three outputs, agricultural eco-efficiency scores are derived by operating MATLAB (R2016a) software. Fig. 3 illustrates the temporal variation characteristics of the average agricultural eco-efficiency of 26 counties in the agricultural production zone of Jilin from 2005 to 2017. During the observation period, the temporal variation characteristics of the average agricultural eco-efficiency values exhibit a flat “M-shaped” fluctuating trend. The agricultural eco-efficiency of the agricultural production zone shows a trend of continuous growth with fluctuation and is characterized by obvious periodic features. Specifically, the average agricultural eco-efficiency first increases steadily with a low growth rate from 2005 to 2010. This is mainly associated with the series of measures implemented to address the three rural issues associated with agriculture, rural areas, and rural peasants and the tax reduction and exemption policies enacted to support the continuous transformation of traditional agriculture to modern agriculture. However, agricultural eco-efficiency then begins to decline with rapid speed from 2011 to 2012, showing a trend of decrease by a wide margin. In this period, the government paid more attention to the “urban disease” accompanying rapid urbanization development, and the counties’ environmental governance in the agricultural sector became looser in the absence of specifically targeted guiding policies for the agricultural sector compared with the urbanization sector. In 2012, the score of agricultural eco-efficiency hits the lowest point of only 0.609. Then, agricultural eco-efficiency bounced back again in 2013–2014, probably because with the implementation of the ecological agricultural production model, the government paid more attention to agricultural resource constraints and strengthened the agricultural policy incentives. In 2015–2016, agricultural eco-efficiency slowly declined again, and it finally increased in 2017. This may be because of the low usage rate of agricultural chemical material, more attention to the agricultural ecosystem, and the high average agricultural eco-efficiency in these years. Overall, the average agricultural eco-efficiency of each year in the agricultural production zone of Jilin was approximately 0.689, varying from 0.609 to 0.766, which is an average level and indicates that there is much room for improvement in agricultural development even though the agricultural eco-efficiency indicates good capacity.

Fig. 3
figure 3

The agricultural eco-efficiency change trend of the JAPZ, 2005-–2017.

Full size image

The spatial distribution characteristics of agricultural eco-efficiency

Fig. 4 plots the boxplot of the agricultural eco-efficiency of 26 counties in 2006–2017. Noticeably, the average agricultural eco-efficiency of Lishu reached 1.32, the highest level, while the values for Yitong, Huadian, Gongzhuling, Nong’an, and Qianguo averaged approximately 1.25, 1.18, 1.05, 1.09, and 1.1, respectively. In addition, the variance of the agricultural eco-efficiency values in Changling, Qian’an, Tonghua, and Dehui is large, which demonstrates that the agricultural eco-efficiency of these counties has large gaps in the efficiency values and the agricultural eco-efficiency is in an unstable state of fluctuation. In contrast, Meihekou, Taobei, Yongji, and Taonan, all of which have weak resource carrying capacity, had very low agricultural eco-efficiency values. For the western counties with the lowest agricultural eco-efficiencies, namely, Taobei and Taonan, this is primarily due to their geographical location with barren saline soil and water shortage, such that they require higher production inputs than other counties. For the central and eastern counties with the lowest agricultural eco-efficiencies, namely, Meihekou and Yongji, the low agricultural eco-efficiency values owe primarily to the inappropriate terrain and incomplete water conservancy facilities. The variance of agricultural eco-efficiency values in Yongji, Jiaohe, Dongfeng, Huinan, Liuhe, Meihekou, Zhenlai, Taonan, Daan, and Taobei is small, but their agricultural eco-efficiency level is low, illustrating the relatively poor stability of the agricultural eco-efficiency level in these counties and the severity of the long-term inefficiency. In contrast, the variance of Nong’an, Huadian, Lishu, Yitong, Gongzhuling, Shuangliao, and Qianguo is small, and their agricultural eco-efficiency level is high, illustrating the relatively good stability of the agricultural eco-efficiency level.

Fig. 4
figure 4

Boxplot of the agricultural eco-efficiency of 26 counties in the JAPZ, 2005–2014.

Full size image

To better observe the spatiotemporal characteristics of agricultural eco-efficiency, this study, based on ArcGIS 10.5 software, categorizes the agricultural eco-efficiency scores calculated by the super SBM method into five levels: low level (0~0.30), medium-low level (0.31~0.6), medium level (0.61~0.90), medium-high level (0.91~1.2), and high level (>1.2). The spatiotemporal distribution map of 2005, 2009, 2013, and 2017 is shown in Fig. 5. As illustrated in Fig. 5, we can view significant spatial differences in agricultural eco-efficiency across counties.

Fig. 5
figure 5

The spatial distribution of agricultural eco-efficiency in the JAPZ in a 2005, b 2009, c 2013, and d 2017

Full size image

In 2005, the high agricultural eco-efficiency values were mainly distributed in counties characterized by a high agricultural economic level, such as Lishu, Gongzhuling, and Changling, with Lishu and Gongzhuling holding the highest ranks. Central counties with a medium-high efficiency level include Qian’an, Qianguo, Nong’an, Dehui, Huadian, Yitong, and Dongliao, while the counties with medium agricultural eco-efficiency are Fuyu, Yushu, and Shuangliao. Other counties characterized by medium-low agricultural eco-efficiency, such as Taobei, Zhenlai, Taonan, Daan, Tonghua, and Yongji, are situated in the western and southeast region; such counties have great potential for improvement. In comparison, central and western counties rely on higher agricultural production input to support agricultural economic growth, which constrains the improvement in agricultural eco-efficiency to some content.

In 2009, Yitong and Lishu County are the counties with a high level of agricultural eco-efficiency, while Qian’an, Qianguo, Nong’an, Dehui, Yushu, Gongzhuling, Tonghua, and Huadian have medium-high efficiency. The distribution of counties at the medium-high level is more concentrated in 2009 than in 2005. The counties with a medium-high level of agricultural eco-efficiency gradually expand from the central to the eastern agricultural production zone in Jilin, and their number increases. In addition, Shuangliao, Dongliao, and Panshi have a medium level of agricultural eco-efficiency, while Dongfeng and Liuhe have low agricultural eco-efficiency. The quantity and magnitude of counties with medium-low agricultural eco-efficiency showed little change.

In 2013, agricultural eco-efficiency decreased slightly, and its spatial agglomeration feature weakened, owing to regional differences in agricultural incentive policies. More than 15 counties had a medium-low level of agricultural efficiency. A possible reason for the decrease in agricultural eco-efficiency is that the execution of agricultural policy weakened. High-level efficiency can be found scattered across the province in Lishu, Yitong, and Huadian, while the counties with medium-high level efficiency are Qianguo, Nong’an, Gongzhuling, Shuangliao, and Panshi. However, the distribution of counties at the medium-high level was more dispersed in 2013 than in 2009. In this year, there was only one county with medium-level efficiency, namely, Yushu County. The counties with medium-low agricultural eco-efficiency were Dongfeng and Tonghua.

In 2017, the county-level agricultural eco-efficiency presents a sharp increase with remarkable spatial variation. Yitong is the only county with a high level of eco-efficiency. The agricultural eco-efficiency in half of the counties is at the medium-high level, showing remarkable club convergence. In particular, the number of counties with agricultural eco-efficiency at the medium-high level gradually increases, and the area of these counties expands from the areas east of Changchun to the central and western areas, while the counties with medium-low agricultural eco-efficiency gradually narrow in quantity and magnitude. The counties with medium-level eco-efficiency are scattered across the counties including Fuyu, Shulan, and Panshi. The number of counties with a medium level of agricultural eco-efficiency in 2017 increases compared with the number in 2013. The values for most counties in the central-southern JAPZ increase from the medium-low level and low level to the medium-high level in 2017. However, there are no longer any counties with low agricultural eco-efficiency.

The four spatiotemporal distribution maps show that the spatial distribution has obvious core-periphery characteristics; that is, the agricultural eco-efficiency of the central counties is generally higher than that of the southeast and northeast counties. In 2005, 2009, 2013, and 2017, there are 6 counties each year—including Nong’an, Huadian, Lishu, Yitong, Gongzhuling, and Qianguo—that achieve high agricultural eco-efficiency. The five counties with agricultural eco-efficiency values higher than 1 exhibit a growth pattern of low inputs, high outputs, and high agricultural eco-efficiency. The counties with a medium-low level of eco-efficiency are concentrated in the southeastern agricultural production zone in Jilin, while those with a low level continue to expand in quantity and are mainly concentrated in the central-eastern agricultural production zone.

Factors influencing agricultural eco-efficiency

Pearson correlation test

A Pearson correlation test between the seven independent variables is conducted before the panel data regression is carried out. Table 4 illustrates the correlation matrices among variables. The test analysis results show that the correlation strength and direction among variables are mostly all significant and that the scores of the correlation matrices are small, which sufficiently confirms that all of the independent variables have weak correlations with each other. Besides, the tested significant level values are less than the critical value of 0.1, which indicates that the variables have passed the two-tailed test. Therefore, the data for these eight independent variables are considered reliable for examining the influencing factors in the 26 counties studied in the panel data regression.

Table 4 Correlation matrices for variables
Full size table

Estimation results of panel data regression

Static panel data models include several main types, namely, the mixed model (MM), fixed effect model (FEM), and random effect model (REM). To determine which model is most appropriate, the F test and Hausman test are required before the construction of the panel data regression model. The F test is applied to determine whether to adopt the MM or the FEM. The Hausman test is applied to determine whether to select the FEM or the REM. The panel data regression model is conducted using the EViews 10.0 software tool. The F test statistic value is 109.594, and its p-value approaches 0.00, which indicates that the null hypothesis of the MM can be rejected, leading us to accept the FEM. Additionally, the test statistic of the Hausman test is 48.837, and its p-value approaches 0.00, which indicates that the null hypothesis of the REM is rejected and likewise suggests the FEM as appropriate. Therefore, the FEM should be established as the appropriate model according to the results of the F test and the Hausman test. The R2 of the FEM and REM are 0.820 and 0.713, respectively, which illustrates that the fit degree of the panel data regression model with fixed effects is better than that of the model with random effects. The FEM regression results of the eight independent variables on agricultural eco-efficiency are shown in Table 5 and are further applied to analyze the influencing factors of agricultural eco-efficiency. The results of the panel data regression set out in Table 5 indicate that some variables, with the exception of the family farmland management scale (FFMS), agricultural industrial structure (AIS), and rural resident income level (RRI), pass the 10% significance level. Specifically, the agricultural technology extension level and agricultural economic development level have positive correlations with agricultural eco-efficiency in the 26 counties during the observation period at the 1% and 10% levels, respectively. Conversely, both the multiple-crop index (MI) and urbanization level (UL) have negative correlations with agricultural eco-efficiency at the 10% and 10% levels.

Table 5 The results of the panel data regression.
Full size table

Analysis of significant influencing factors

The estimated coefficient of multiple-crop index (MI) is negative and significant at the 1% level, implying that multiple-crop index suppresses the improvement in agricultural eco-efficiency as a whole. On the one hand, this finding is associated with the 1-year farming system, which further decreases the negative environmental impact in the agricultural production process. On the other hand, specifically, JAPZ is located in the cold latitude zone, and the climate heat is insufficient, which by the way leads to low-frequency farming activities. Less frequent farming activities will lead to less agricultural carbon emissions with the consequence that the multiple-crop index promotes the improvement in agricultural eco-efficiency. That is, a lower multiple-crop index means less farming activities, which is beneficial for agricultural eco-efficiency. Therefore, when the multiple-crop index is low, the agricultural eco-efficiency is large.

Additionally, the correlation coefficient between the agricultural technology extension level (ATEL) and agricultural eco-efficiency is positive and significant at the 1% level, indicating that improvement in the agricultural technology extension level tends to intensify the increase in agricultural eco-efficiency. This is because agricultural professionals and technical personnel can guide farmers to implement environmentally friendly agricultural production methods, which is conducive to driving productivity and optimizing the production process via the technology effect. In addition, as displayed in Table 5, the estimated correlation coefficient is 0.058 and passes the significance test, which demonstrates that this indicator can sufficiently influence agricultural eco-efficiency, but the effect is not obvious. This result directly confirms that environmental technology and environmental management skills can bring about an improvement in overall agricultural eco-efficiency. That is, counties with more agricultural technology support for agriculture have the ability to curb negative impacts on the agricultural ecological environment.

The correlation coefficient of agricultural economic development level (AEDL) and agricultural eco-efficiency positively pass the significance test at the 1% level, indicating that an increase in agricultural economic development level can accelerate the increase in agricultural eco-efficiency. Numerically, the correlation coefficient of AEDL, reaching 0.084, is larger than the correlation coefficients of the agricultural technology extension level variable, which illustrates that the agricultural economic development level occupies the leading position in the development of agricultural eco-efficiency and the “economic attributes” are important for agricultural eco-efficiency. A possible reason for the positive effect of agricultural economic development level on agricultural eco-efficiency is that there is large number of national agricultural counties in the JAPZ. The obvious agricultural development scale effect can not only stimulate the expansion of agricultural production but also improve farming methods and optimize agricultural materials for the high-quality development of agricultural production. In addition, the counties with a high agricultural economic development level are likely to achieve a balance between agricultural production and the ecological environment with the growth of agricultural intensification and specialization, which is conducive to increasing agricultural eco-efficiency.

The correlation coefficient of urbanization level (UL) indicates that this indicator has a significant negative influence on agricultural eco-efficiency at the 1% level, which indicates that the increase in urbanization level can hinder agricultural eco-efficiency. Moreover, the negative correlation coefficient of urbanization level is lower than that of multiple-crop index in terms of the impact degree. This finding illustrates that among the selected variables, the urbanization level featuring socioeconomic attributes is the hindrance to agricultural eco-efficiency. Since the eleventh five-year plan period, surplus labor has flowed between urban and rural areas, accompanying rapid urban construction. As a consequence of the surplus labor transfer to urban areas, urban development has also squeezed out the input of labor, capital, and other factors required for agricultural production, leading to changes in the employment structure. The laborers who remain in the countryside must use more agricultural machinery to compensate for the loss in labor via the substitution effect, which can cause the deterioration of the agricultural ecological environment. It is not difficult to understand that when the urbanization level increases, the agricultural eco-efficiency may ultimately decrease.

Robustness test of panel data regression model

To verify the reliability of the empirical results, this paper re-estimates the model by applying the method of changing the observation period and divides the observation period into two parts, 2005–2011 and 2012–2017, to test the empirical results of the factors affecting agricultural eco-efficiency in the JAPZ. The detailed empirical regression results are shown in Table 6. Although the coefficients and significance of some independent variables have changed, the basic conclusions are still similar to the abovementioned full-sample regression results. Additionally, to further examine the robustness of the regression results, this paper applies the method of replacing independent variables to re-estimate the model. When conducting the robustness test, the urban population percentage in the total population is replaced by the proportion of the tertiary industry in GDP and other independent variables remain unchanged, and then the panel regression estimation is performed again. The regression results are also displayed in Table 6. The regression coefficient of the proportion of the tertiary industry in GDP has a significant and positive influence on agricultural eco-efficiency at the 5% level, which is contrary to the regression coefficient of the urban population percentage in the total population. This finding suggests that the development of the tertiary industry can extend the agricultural industry chain and increase the added value of agricultural products, thereby promoting agricultural eco-efficiency. The estimated coefficients of other independent variables, such as the multiple-crop index, agricultural economic development level, and agricultural technology extension level, are consistent with the aforementioned regression results in the influencing direction. In summary, we can consider the regression results to be robust. Finally, considering that there may be missing variables or two-way causal relationships between variables in the model construction resulting in endogenous issues and leading to instability of the model’s regression results, this paper applies the lagging dependent variable and lagging agricultural economic development level variable instrumental variables through the two-stage least squares (TSLS) method. The significance and impact properties of each explanatory variable are basically consistent with the abovementioned regression results.

Table 6 Robustness test of the regression model.
Full size table

Conclusions and policy suggestions

Conclusions

During the observed period of 2005–2017, the average value of agricultural eco-efficiency in the JAPZ exhibited a rising trend with fluctuations. The agricultural eco-efficiency of each county could improve in terms of resource conservation and environmental protection, although it is still at a medium level overall. This study also reveals that there are significant spatial differences in agricultural eco-efficiency in the JAPZ. The agricultural eco-efficiency in the region presents a spatial pattern that progresses from the core to the periphery. Specifically, central counties usually have higher agricultural eco-efficiency than southeastern and northwestern counties. Nong’an, Huadian, Lishu, Yitong, Gongzhuling, and Qianguo achieved high agricultural eco-efficiency in 2005, 2009, 2013, and 2017. Since the change in agricultural eco-efficiency results from the complex interaction of various factors, this study further considers various influencing factors that lead to the change in agricultural eco-efficiency. The panel data regression estimation results indicate that the agricultural technology extension level, multiple-crop index, agricultural economic development level, agricultural technology extension level, and urbanization level have close correlations with agricultural eco-efficiency. The agricultural economic development level occupies the leading position in the development of agricultural eco-efficiency, while the multiple-crop index presents the greatest hindrance to agricultural eco-efficiency.

Policy suggestions

According to the aforementioned contributing factors, suggestions are proposed for policy makers. First, the results show that multiple-crop index constrains agricultural eco-efficiency. The multiple-crop index is an obvious factor that features natural environmental characteristics. Governments in the JAPZ should pay more attention to the intensity and scale of agricultural farming activities and adopt new environmentally friendly behavior to maintain the stability of the agricultural ecological environment. Second, the correlation coefficient of agricultural technology extension level is not high, but it is large enough to have an impact on agricultural eco-efficiency. This indicates that its key role in the improvement in agricultural eco-efficiency should not be neglected, although the agricultural technology extension level exerts a slight effect on agricultural eco-efficiency. Technical training and professional skills should be provided for farmers in the JAPZ to help them better master relevant green and sustainable agricultural technologies. Third, among the selected variables, the agricultural economic development level is the strongest positive factor driving agricultural eco-efficiency, implying that the scale expansion and total growth of the agricultural economy are still the key ways to promote agricultural eco-efficiency. Hence, the continuous improvement of the agricultural economic development level is one of the vital ways to increase agricultural eco-efficiency. Taking ecological priorities and green development as guidance, the extensive agricultural production and management model should be transformed, the development methods should be optimized, circular and ecological agriculture should be developed, and the sustainable use of agricultural resources should be promoted to support high-quality agricultural development in the JAPZ. Finally, the urbanization level exerts a negative impact on agricultural eco-efficiency, indicating that ecology-oriented agricultural subsidies should be improved and a high-efficiency compensation mechanism should be established to stimulate enthusiasm for agricultural production in the JAPZ.

There is an urgent practical need for research on agricultural eco-efficiency under resource and environmental constraints. This study aims to narrow the gap in the literature on agricultural eco-efficiency, but several limitations remain that deserve in-depth attention in future research. In fact, agricultural eco-efficiency is also affected by natural factors, such as climate, soil properties, and the natural environment, which influence the input and output of agricultural production to some extent. Due to the limited availability of data regarding natural factors, this study investigated only socioeconomic factors. When more natural data are publicly available, the spatial dimension and the indicators applied in our study could be improved upon to enable more comprehensive modeling. Moreover, the influencing factors of agricultural eco-efficiency could be further investigated from the perspective of national agricultural production zones to help accomplish China’s future carbon emission reduction targets ahead of schedule and achieve the green agricultural transition.