Impacts of the integral development of agriculture and tourism on agricultural eco-efficiency: a case study of two river basins in China

Abstract

The integral development of agriculture and tourism heavily depends on natural resources, especially agricultural resources, and it also may contribute to the sustainability of agriculture. The sustainability of agriculture is usually measured by the agricultural eco-efficiency. The main objective of this paper is to demonstrate the impact of the agriculture and tourism integration on agricultural eco-efficiency, so as to reflect the role of the agriculture and tourism integration on promoting agricultural sustainability. Applying panel data of 19 provinces in Yangtze River Basin and Yellow River Basin in China from 2008 to 2019, this paper uses a PSTR model to demonstrate the impact of agriculture and tourism integration on agricultural eco-efficiency and its mechanisms. It is found that in the study area, the integration of agriculture and tourism can contribute to improvements in agricultural eco-efficiency, and when levels of integration are high, the promotional effect is enhanced. However, impacts vary among different regions to some extent. In terms of influencing mechanisms, inputs of agricultural labor, sown crop area, effective irrigation area and other factors decrease with continuous improvements of integration. The input of pesticides, fertilizers and other factors does not decrease in the initial stage of integration but undergoes a rapid decrease after crossing a threshold value. And it demonstrates that the integration of agriculture and tourism is beneficial for improving the agricultural eco-efficiency, so that it will contribute to agricultural sustainability. Meanwhile, this paper offers a new research perspective for discovering effects of agriculture and tourism integration, providing implications for formulating regional policies for the integration.

1 Introduction

China is a big agricultural country with the most largest population in the world, which has made remarkable achievements in agricultural industry (Deng & Gibson, 2019). From 1978 to 2019, the total output of grain, aquatic products and meat in China increased by several times, accounting for 1/5, 1/3 and 1/4 of the world’s supply of these products (Liu et al., 2019; Zeng et al., 2021). The rapid growth of China’s agriculture is mainly due to the continuous increase in input factors, which also leads to serious agricultural endogenous pollution. In terms of input factors, the use of pesticides, fertilizers and agricultural plastic films increased by 2.2 times, 6.7 times and 5.2 times, respectively (Liu et al., 2020a; Shi, 2022). The effective utilization rates of chemical fertilizer and pesticide are very low; the rate of the former is less than 1/3, and the rate of the latter is less than 2/3. Agricultural growth in China mainly depends on the intensive input of production factors. The pattern of agricultural production has led to serious agricultural pollution (Liu et al., 2019; Zou et al., 2020). And it has replaced industrial pollution as the biggest source of pollution in China (Zhao et al., 2022). In recent years, with aggravation of negative effects of agricultural production and the upsurge of sustainable development all over the world, to protect agricultural environment and achieve sustainability in agriculture has become an important issue concerned by the government and academia. Therefore, agricultural eco-efficiency has become a hot topic of research, by which agricultural sustainability can be reflected.

Technological progress and urbanization have stimulated people’s demand for tourism, among which agriculture-related tourism is an adequate way to meet people’s needs, which can provide a relaxed and comfortable leisure environment (Ćirić et al., 2021). As a result, multi-functions of agriculture besides providing agricultural products have been explored, such as its function of tour, leisure and entertainment. Therefore, based on agricultural industry chain and featured by agriculture-related tourism, various forms of agriculture and tourism integration have gradually taken shape (Deng et al., 2019), such as agri-tourism, leisure agriculture and rural tourism (Karampela et al., 2021). The Guideline on Comprehensively Promoting Rural Revitalization and Accelerating Agriculture and Rural Modernization in China clearly states that the overall development of agriculture and tourism is an important way to promote rural revitalization and deepen supply-side structural reform. By the end of 2020, China had set up 388 demonstration counties as leisure agriculture and rural tourism destinations, and promoted 1,216 “China’s Beautiful Leisure Villages” and 1,199 “National key villages for rural tourism” (Wang et al., 2022). According to data from the Ministry of Agriculture in China, the number of leisure agricultural tourists is 1.5 billion tourists, which generated revenue of 360 billion Yuan in 2014, while the number of leisure agricultural tourists increased to 3.2 billion, which generated revenue of 850 billion Yuan. In 2019, the total number of leisure agricultural tourists accounted for 53.28% of all the domestic tourists, while the operating income it generated accounted for 14.83% of the operating income of domestic tourism in China, which can be seen in Table 1.

Table 1 Development of leisure agriculture and rural tourism in China

With agriculture and tourism integration, its comprehensive effect can be further highlighted. On one hand, the integration of agriculture and tourism plays an significant role in solving bottlenecks of agricultural and tourism development (Schilling et al., 2012). On the other hand, it can effectively extend and expand the industry chain of agriculture, promote development of agro-related industries, and provide more jobs for farmers, thereby increasing their incomes (Veeck et al., 2016). A leisure farm with an annual reception of 100,000 people can achieve a business income of 10 million Yuan and increase the income of more than 1,000 agricultural families.Data from China National Leisure Agriculture and Rural Tourism Parks show that the sales revenue of agriculture and related products in the past five years is equivalent to the total operating revenue of the parks. According to the data of 1,000 key rural tourism villages in the first quarter of 2021 in China, the contribution of rural tourism employment is 47.1% on average (Tourism Big Data Center of Ministry of Culture and Tourism in China 2021). The agro-tourism activities is deeply affected by ecological, economic and tourism resources, of which ecological factors are the most critical (Koscak, 1998; Lupi et al., 2017; Torres, 2003; Wei, 2015). Agriculture-related tourism activities, such as harvesting local products and gathering forest fruits, also may affect the environment (Ammirato et al., 2020; Ćirić et al., 2021), especially on agriculture environment. However, few research and data have focused on the integral development of these two industries on agricultural environment or agricultural sustainability.

Against above context, main purposes of this paper are as follows: 1) to measure the integral level of agriculture and tourism to better identify the linkage of agriculture and tourism, 2) to assess the agricultural ecological efficiency involving multi-outputs and multi-inputs, 3) to demonstrate impacts of the agriculture and tourism integration on agricultural eco-efficiency, and 4) to identify influencing mechanisms of agriculture and tourism integration on agricultural eco-efficiency so as to put forward some policy recommendations. The contributions of this study are as follows: Firstly, it demonstrates the impact of the integral development of agriculture and tourism on agricultural eco-efficiency with empirical analysis, providing a new perspective for exploring factors which may affect agricultural eco-efficiency and agricultural sustainability; secondly, it focuses on the environmental effect of the agriculture and tourism integration from the perspective of impacts on agricultural eco-efficiency, while most studies concerning its economic and social impacts.

2 Literature review

2.1 The integral development of agriculture and tourism

2.1.1 Connotation of integral development of agriculture and tourism

CAMPBELL et al. (2007) believed that an industry could cause changes in other industrial structures through inter-industrial linkage, providing possibilities for inter-industrial coupling development Tourism industry is a highly comprehensive and highly correlated industry, which has the possibility of integration with agriculture and its related industries. Tourism revenue determines the demand for tourism products and then determines the production of agricultural products and food in the process of tourism consumption. (Ristić et al., 2019; Rogerson, 2012). On the other hand, agricultural production process and seasonal characteristics affect the content of tourism supply (Sanches-Pereira et al. 2017; Torres, 2003). Given the strong linkage of tourism and agriculture, increasing attention has been given to the integration of agriculture and tourism (Gilbert & Hudson, 2000; Ristić et al., 2019; Streifeneder, 2016). Nie and Fang (2019) believes that the integration of agriculture and tourism is a process in which the internalization of inter-industrial division of labor, blurriness of boundary, sharing of products, markets and resources and the symbiosis development of the two industries are driven by market demand, economic growth and competition. Agriculture and tourism integration refers to the process of regional agro-tourism resource development, agricultural production and marketing, spatial layout and management implemented by the local government, agro-tourism enterprises, farmers and other relevant stakeholders to maximize economic, social and ecological benefits under specific local environment and consumption conditions (Wang, 2018). In terms of its nature, agricultural and tourism integration is a typical cross-type integration mode. In terms of its nature, the integration of agriculture and tourism is a typical cross-type integration mode. It is the application of leisure, experience, sightseeing and other agricultural functions in tourism industry chain, and also the play of tourism and leisure functions in the agricultural industry chain. In the integration of agriculture and tourism, the innovation result of agriculture changing the tourism industry chain is manifested as the production of tourism and leisure agricultural products and the new industrial forms of leisure, sightseeing and experience agriculture gradually formed based on this. The innovation results of tourism industry to change the agricultural industry chain are mainly reflected in the mining of tourism resources in agricultural production. This kind of agricultural products based on cross-integrated industrial resources expands the types of existing agricultural products and also changes the functions of the original products of these industries. At the same time, the supporting services related to agriculture are gradually derived to expand the versatility of agriculture, which can also be regarded as the integration mode of agricultural function expansion, which is shown in Fig. 1. Based on the cross-penetration between agriculture and tourism industry, the value creation function of tourism industry is gradually penetrated into the agricultural system, and the agricultural versatility is implanted in the tourism industry system, thus forming a new industrial form. Its essence is the deconstruction and extension of the value chain of agriculture and tourism industry, which can drive the integration, optimization and reorganization of resources, factors, technologies and market demands in rural areas, so as to enrich the connotation of agriculture, diversify the forms of industrial expression and diversify the forms of farmers’ income.

Fig. 1

Conceptual frame of the integral development of agriculture and tourism R&D is short for research and development

2.1.2 Measurement of the integration level of agriculture and tourism

Scholars have evaluated the level of industrial integration from the perspective of industrial association or industrial interaction. Studies from this perspective focus more on the evaluation of the state of interactive integration between industries. Similarly, scholars carrying out agriculture and tourism integration research usually measure the integration level of agriculture and tourism based on the correlation degree, coordination degree and industrial interaction between agriculture and tourism industry. Commonly used measurement methods from this perspective mainly include input–output method (Ye et al., 2018), coupling model (Wang, 2018; Zhang & Zhang, 2020; Zhou et al., 2016), gray correlation analysis method (Zhang & Zhang, 2020), co-integration analysis method (Nie & Fang, 2019), analytic hierarchy process method (Yang et al. 2021b; Chen et al., 2016) and vector autoregressive model (Xia and Xu 2016; Su, 2017; Liu et al., 2020a).

2.1.3 Effects of agriculture and tourism integration

As for the effects of the integration of agriculture and tourism, lots of existing papers have focused on its economic, social and environmental effects. Some scholars believe that the interaction and integration of agriculture and tourism can create new and broad market space and expand consumer demand, while also promoting the high-quality development of agricultural products and tourism (Asiedu et al. 2011; Wang et al., 2022; Ilbery, 1991). And it improves the quality of agricultural products to better meet the needs of tourists and create more employment opportunities, which can increase farmers’ income, thereby reducing poverty and promoting economic development (Adiyia & Vanneste, 2018; Comerio & Strozzi, 2019; Pitrova et al., 2020; Veeck et al., 2016). At the same time, it makes the surplus rural labor force engaged in tourism reception work in their spare time, which is conducive to solving the shortage of local tourism labor force; the development of tourism expands the demand for agricultural products and can solve the problems of insufficient market demand for agricultural products (Podovac et al., 2019; Rogerson, 2012; Schilling et al., 2012). And it is also beneficial to rural environment (Podovac et al., 2019; Sanches-Pereira et al. 2017). On the whole, the integration gives full play to the versatility of agriculture, explores the resource potential of agriculture and rural areas, and realizes the economic benefits of agricultural ecological resources in rural areas. On the other hand, the sustainable development of agricultural and tourism integration heavily depends on protection and sustainable utilization of these resources. Therefore, the integration of agriculture and tourism also forces the protection of agricultural ecological environment in turn.

2.2 Agricultural eco-efficiency

2.2.1 Concept of agricultural eco-efficiency

In 1990, Schaltegger and Sturn put forward the concept of ecological efficiency (eco-efficiency) in academia. They defined eco-efficiency as the ratio between economic growth and environmental impact. High eco-efficiency means maximum productivity with the least resource consumption, and minimizing environmental stress (Keating et al., 2010; Qu et al., 2022; Rosano-Peña et al., 2021). It was not until 1998 that the Organization for Economic Cooperation and Development (OECD) initiated the use of eco-efficiency in agricultural sector, in order to carry out effective assessments of agricultural production and its environmental impact (OECD, 1998). Since then, there have been more and more literatures on agricultural eco-efficiency, and great achievements have been made (Liu et al., 2019; Maia et al., 2016; Pishgar-Komleh et al., 2021). According to connotation of eco-efficiency, agricultural eco-efficiency (AEE) refers to the minimization of agricultural production factor input and undesired output under the given condition of expected output (Picazo et al. 2011).

2.2.2 Measurement of agricultural eco-efficiency

Some studies have assessed AEE based on the broad sense of agriculture, which contains the farming, husbandry, forestry, and fishing (Toma et al., 2015; Pan et al. 2013; Hong et al., 2016; Liu et al., 2020b). Other studies focused on the narrow sense of agriculture, which means crop plantation only (Wang et al. 2016;Wang et al. 2021;Xu et al., 2014; Shang et al., 2020;Picazo et al. 2011). In terms of non-expected output indicators, agricultural non-point source pollution (Zou et al., 2020), agricultural carbon emission (Pang et al., 2016; Liu et al., 2020b; Tian 2014; Li 2014) and agricultural gray water pollution (Wang et al. 2021)are commonly used indicators. Based on AEE measurement, spatial–temporal evolution characteristics of AEE have been analyzed (Liu et al., 2020b). To evaluate agricultural ecological efficiency, many methods were constructed, such as Life Cycle Assessment, Ratio Method (Lyrstedt, 2005), Stochastic Frontier Analysis (SFA) (Bravo-Ureta & Pinheiro, 1993; Deng & Gibson, 2019; Jin et al., 2018; Suzigan et al., 2020) and Data Envelopment Analysis (DEA) (Alene & Zeller, 2005;Pang et al., 2016; Picazo et al. 2011; Toma et al., 2015). Among them, DEA and SFA are the two main methods, which can achieve convincing results in most cases.

2.2.3 Influencing factors of agricultural eco-efficiency

As for the influencing factors of AEE, some studies have focused on factors of agriculture itself, such as the agricultural scale level, agricultural industrial structure, per capita agricultural added value, agricultural disaster rate, and agricultural machinery density (Pang et al., 2016; Rybaczewska-Błazejowska & Gierulski, 2018), while others analyzed relevant external influencing factors of agriculture, such as educational level of agricultural labor force, farmers’ household operating income ratio, public investment in agriculture, government regulation and technological progress (Liu et al., 2020b; Wang & Zhang, 2018). The literature review shows that previous studies have paid less attention to the influencing factors of AEE from the perspective of industrial integration.

2.3 Impacts and its mechanism of agriculture and tourism integration on AEE

The agriculture and tourism integration is based on agriculture, which is the extension of agricultural multi-function and industrial chain. So agriculture-related tourism activities are largely depend on agricultural resources, which also influence the agriculture environment (Ćirić et al., 2021). First, the exchange of information and elements between agriculture and tourism will improve the quality of agricultural labor on one hand and transfer a number of surplus rural labor to cities on the other hand (Dax et al., 2019; Liu et al., 2014). The improvement of agricultural labor quality effectively improves AEE, and the increase in the number of labor transferred to other sectors reduces the input of agricultural labor, thus improving the level of agricultural effective output (Forsyth, 1995). Second, the integration of agriculture and tourism is helpful to spread advanced agricultural technologies. The agriculture-related leisure business is helpful to realize the ecological environment value of agriculture (Liu et al., 2014; Pitrova et al., 2020; Wang et al., 2022), which helps to improve the capital accumulation level of agricultural producers and provides them with adequate capitals to purchase advanced and efficient agricultural machinery and equipment. As a result, it reduces the input of labor, land, mechanical power and other factors, which is beneficial to improve AEE. In addition, in order to promote industries integration, the government always takes advantage of its own resource advantages to protect rural arable land resources, revitalize idle rural arable land and create preconditions for the use of advanced agricultural technology (Yin & Xiao, 2019), so that agriculture input factors such as water and arable land can be saved a lot, and AEE is improved ultimately. Thus, agriculture and tourism integration is beneficial to reduce the agriculture inputs, especially those are harmful to environment (Sanches-Pereira et al. 2017). Meanwhile, the integration can strengthen the management of agricultural ecological environment. And it has gradually expanded the scale of non-agricultural industries, which enables government departments with enough financial support to improve the agricultural production environment (Colton & Bissix, 2005), such as promoting construction of saving water facilities and the education of ecological protection, so that ecological capital can be accumulated, which is the foundation for improving of AEE (Bayyurt & Yılmaz, 2012; Chakraborty, 2021). At the same time, in order to promote the sustainable development of agriculture and tourism integration, governments have continuously strengthened regulations on the usage of chemical fertilizers or pesticides, which results in reduction of these chemical products (Deng et al., 2019; Xie et al., 2018). On the other hand, the integration helps to strengthen the ecological behavior of agricultural producers (Lee & Jan, 2019). With the integral development, agricultural producers gradually begin to realize that agro-environmental ecological factors can provide them with create higher and sustainable premiums (Barbieri, 2013; Jiang, 2022). In order to achieve long-term sustainable economic benefits, they will set up green development concept, practice and strengthen environmental protection behavior, reducing harmful environmental elements inputs, thus promoting AEE (Setokoe et al., 2019). The influencing mechanism of agriculture and tourism integration on AEE is shown in Fig. 2.

Fig. 2

The impact mechanism of agriculture and tourism integration on AEE

However, few empirical studies have concentrated on the influence of agriculture and tourism integration on AEE. Given the contemporary situation in China and the above analysis, it was hypothesized that: The integration of agriculture and tourism might contribute to improving the agricultural eco-efficiency, and this effect will be enhanced when the level of integration of agriculture and tourism increases.

3 Methods and materials

3.1 Evaluation of the integral level of agriculture and tourism based on the synergetic degree model

Industrial integration can be regarded as a complex and changing open system formed through competition and cooperation among different industrial components in two or more industrial systems in time and space (Rui and Hu 2008). An industrial integration system is a self-organized system. Based on synergetic self-organization theory (Liening, 2014), this paper uses the synergetic degree model of the composite system proposed by Tao et al. (2017) to estimate the integrated level of agriculture and tourism. Suppose that the subsystem of the composite system of agriculture and tourism is S_i, i ∈ [1, 2], and the order parameter in its development process is e_i = (e_i1,e_i2,…,e_in) where n ≥ 1, β_ij ≤ e_ij ≤ α_ij, j = 1,2,…,n. In the formula (2), α_ij is the upper limit of order parameter component e_ij and β_ij is the lower limit (Huang et al. 2020). For the lower limit, the minimum actual value for the past year can be used. The upper limit can be selected according to its predicted value, planned value, or maximum actual value in the past year. Then, the order degree of e_ij of the subsystem can be calculated as follows:

$$\mu_{i} \left( {e_{ij} } \right) = \left\{ {\begin{array}{*{20}c} {\frac{{e_{ij} - \beta_{ij} }}{{\alpha_{ij} - \beta_{ij} }},} & {j \in \left[ {1,\;k} \right]} \\ {\frac{{\alpha_{ij} - e_{ij} }}{{\alpha_{ij} - \beta_{ij} }},} & {j \in \left[ {k + 1,\;n} \right]} \\ \end{array} } \right.$$

(1)

According to the above equation, μ_i(e_ij) ∈ [0, 1], the larger μ_i(e_ij) is, the greater the “contribution” of order parameter component e_ij to the system order(Wan et al., 2021). We must note here that in a real system, there are a number of e_ijs that are either too high or low, but it is better to concentrate around a particular point. For an e_ij value that does not meet the order degree definition, it can be satisfied by adjusting its value range [β_ij, α_ij]. The order degree not only depends on the number of order parameters, but it is also closely related to their combination (Qiao et al., 2017). In this paper, the linear weighted summation method (Jiao et al., 2018) is used to calculate the order degree of the subsystem, which can be written as Eq. (2):

$$\mu_{i} \left( {e_{i} } \right) = \sum\limits_{j = 1}^{n} {\lambda_{j} \mu_{i} \left( {e_{ij} } \right),\lambda_{j} \ge 0} ,\;\sum\limits_{j = 1}^{n} {\lambda_{j} = 1}$$

(2)

μ_i(e_i) is the system order degree of subsystem S_i. The larger μ_i(e_i) is, the more e_i contributes to subsystem Si and vice versa (Wan et al., 2021). In Eq. (4), λ_j represents the weight value corresponding to each order parameter component (Amedzro St-Hilaire & Boisselier, 2019). Here, the method proposed by Diakoulaki (1995) is used to calculate the weight of the subsystem evaluation index. For original data, it is assumed that each order parameter component includes m samples for each year, among which x_jt is the value of the jth component of the order parameter t sample. The original data x_jt are standardized as follows:

$$x_{jt}^{\prime } = {{\left( {x_{jt} - \overline{x}_{jt} } \right)} \mathord{\left/ {\vphantom {{\left( {x_{jt} - \overline{x}_{jt} } \right)} {S_{jt} \left( {j = 1,\;2,\; \cdots ,\;n;\;t = 1,\;2, \cdots ,\;m} \right)}}} \right. \kern-\nulldelimiterspace} {S_{jt} \left( {j = 1,\;2,\; \cdots ,\;n;\;t = 1,\;2, \cdots ,\;m} \right)}}$$

(3)

where $x_{jt}^{\prime }$ represents the standardized data, $\overline{x}_{jt} = \frac{1}{m}\sum\limits_{m = 1}^{m} {x_{jt} }$ is the mean value of variable, and S_j is the standard deviation of variable. We then estimate the weight of subsystem evaluation index with the CRITIC method.

The basic principles of CRITIC:

$$\xi_{j} = \sigma_{j} \cdot \sum\limits_{j = 1}^{n} {\left( {1 - r_{jk} } \right)\left( {j = 1,\,2, \cdots ,n} \right)}$$

(4)

In Eq. (4), ξ_j reflects the influence of the jth evaluation index on the system of a given province, σ_j is the overall standard deviation of the evaluation index after standardization, and r_jk is the correlation coefficient between the jth and kth evaluation indexes. The higher ξ_j is, the stronger is the influence of the jth index on the system, the greater the index’s relative importance. λ_j is the weight of the jth evaluation index, which can be written as:

$$\lambda = \frac{{\xi_{j} }}{{\sum\limits_{j = 1}^{n} {\xi_{j} } }}\left( {j = 1,\;2, \cdots ,n} \right)$$

(5)

System synergy not only involves the orderly development of the subsystem itself but also coordination between subsystems. According to the theory of the systemic coordination, considering the synergetic status of two systems, it can be measured by the deviation between them. In this paper, the deviation coefficient C is introduced. The smaller the deviation coefficient is, the better the coordination between systems will be, and conversely, the worse the coordination between systems will be. Assuming that the order degrees of subsystems S₁ and S₂ are μ₁(e₁) and μ₂(e₂), respectively, the synergy of the two subsystems can be described as follows:

$$C = \sqrt[2]{{\frac{{\mu_{1} \left( {e_{1} } \right)\;\mu_{2} \left( {e_{2} } \right)}}{{\left( {\mu_{1} \left( {e_{1} } \right) + \mu_{2} \left( {e_{2} } \right)} \right)^{2} }}}}$$

(6)

$$F = {{\left( {\mu_{1} \left( {e_{1} } \right) + \mu_{2} \left( {e_{2} } \right)} \right)} \mathord{\left/ {\vphantom {{\left( {\mu_{1} \left( {e_{1} } \right) + \mu_{2} \left( {e_{2} } \right)} \right)} 2}} \right. \kern-\nulldelimiterspace} 2}$$

(7)

In Eq. (7), F is the comprehensive harmonic index of subsystems of agricultural and tourism, reflecting the overall contribution of the two subsystems (Wan et al., 2021). To reflect overall “coordination” or “synergy” effects of subsystems of agriculture and tourism, the synergy model of the composite system is constructed based on the synergy model of the two subsystems as follows:

$$D = \sqrt {C \times F}$$

(8)

In Eq. (8), D is the synergy degree of the composite system, which presents the integration level of two industries. So we use it to measure the integration level of agriculture and tourism(ATL).

The agriculture and tourism integration industry subsystems involve several system state variables. According to synergetic self-organization theory (Liening, 2014), order parameters can determine the macroscopic behaviors of a system and represent its order degree (Li et al., 2021). These parameters can be referred to as ordering parameters. Therefore, accurately determining a system’s order parameter can not only effectively reduce the complexity of system research but can also capture the development of composite system evolution.

According to the existing literature, 7 indexes (the primary industry output value, gross output value of agriculture, forestry, husbandry and fishery, consumption level of rural residents, output value of featured agricultural products, crop sown area, number of people employed in primary industry, total power of agricultural machinery) are selected for agriculture industry (Liu et al., 2020b; Wang, 2018) as order parameters to measure levels of agriculture development according to the performance and features of the agricultural industry. 7 tourism indexes are selected as order parameter components to measure tourism development (domestic tourism revenue, foreign exchange earnings from tourism, the number of domestic tourism, the number of inbound travelers, the number of star hotels, the number of scenic spots above grade A, the number of people employed in tourism) according to the performance and features of tourism industry. All the above indicators and their description are given in Table 2.

Table 2 Variables and indicators description of agriculture and tourism integration

3.2 Assessment of agriculture eco-efficiency based on Super-SBM model

Agricultural production will bring desired output, at the same time bring undesirable output, such as various of pollutants (Liu et al., 2020b; Rosano-Peña et al., 2018). In this paper, the Super-SBM model with undesirable output is used to measure AEE for the following two reasons: First, as agricultural production inputs such as fertilizers and pesticides destroy the ecological environment, the measurement of agricultural output should consider the negative impact of agricultural production on the ecological environment at the same time. Second, in the ordinary SBM model, the DMU whose efficiency level is greater than or equal to 1 is assigned a value of 1, so the DMU whose efficiency value is greater than 1 cannot be distinguished from the DMU whose efficiency value is 1(She et al., 2020; Srinivasa Raju & Nagesh Kumar, 2013). However, the super-efficiency SBM model including undesired output directly incorporates the slack variables into the objective function, thus avoiding the above problem (Xiang et al., 2019). The super-efficiency SBM model with unexpected output is given as follows:

$$\left\{ {\begin{array}{*{20}l} {\rho * = \min \frac{{1 + \frac{1}{m}\sum\limits_{i = 0}^{m} {D_{i}^{ - } } }}{{1 - \frac{1}{a + b}\left( {\sum\limits_{r = 1}^{a} {\frac{{D_{r}^{e} }}{{y_{rk}^{e} }}} \sum\limits_{h = 1}^{b} {y_{hk}^{f} } } \right)}}} \\ {s.t.\sum\limits_{j = 1,\;j \ne k}^{n} {x_{ij} \lambda_{j} \le x_{ij} + D_{i}^{ - } } } \\ {\sum\limits_{j = 1,\;j \ne k}^{n} {y_{rj}^{e} \lambda_{j} \le y_{hk}^{f} } } \\ {\sum\limits_{j = 1,\;j \ne k}^{b} {y_{hk}^{f} \lambda_{j} \le y_{hk}^{f} + D_{h}^{f} } } \\ {1 - \frac{1}{a + b}\left( {\sum\limits_{r = 1}^{a} {\frac{{D_{r}^{e} }}{{y_{{}}^{{}} }}} + \sum\limits_{h = 1}^{b} {\frac{{D_{h}^{f} }}{{y_{hk}^{f} }}} } \right)} \\ {D^{ - } \ge 0,\;D^{e} \ge 0,\;D^{f} \ge 0} \\ \end{array} } \right.$$

(9)

Assuming that λ ≥ 0 and ∑λ = 1, this is the super-SBM model of variable scale return (VRS) with unexpected outputs. Assuming that λ ≥ 0, this is the supper-SBM model of constant return to scale (CRS) with unexpected outputs. Here, D⁻, D^ε, and D^f are slack variables, respectively, representing too much input, insufficient expected output and too much unexpected output. ρ* is the eco-efficiency of a DMU, and since the model is super-efficient, eco-efficiency is no longer between 0 and 1.

Crop plantation is still the main part of agricultural industry in China, so this study takes agriculture in the narrow sense as the research object and selects input and output indicators to evaluate AEE. The output indicator includes expected and unexpected outputs. The total agricultural output is selected as the expected outputs. Carbon emission from agricultural production is selected as the unexpected output (Pang et al., 2016), which originate from the use of fertilizers, pesticides, agricultural irrigation and powered agricultural machinery. According to West and Marland (2002), the carbon emission coefficients of the four categories are as follows: pesticides 4.93 (kg/kg), chemical fertilizers 0.90 (kg/kg), agricultural irrigation 20.48 (kg/ha), and the total power of agricultural machinery 0.18 (kg/kw) (Chen et al., 2020; Liu et al., 2021). The unexpected outputs calculation is as follows:

$$E = \sum {T_{i} \times \delta_{i} }$$

(10)

E represents the total carbon emissions, T means the source of carbon emissions, δ is the carbon emission coefficient (Liu et al., 2021), and i represents the ith carbon emission source (Han et al., 2018).

Referring to existing papers on AEE (Keating et al., 2010; Liu et al., 2020b; Pang et al., 2016), 5 input variables are selected: (1) agricultural labor, which is measured by the “agricultural labor force” (Liu et al., 2020b); (2) land, which is measured by the total sown agricultural area (Liu et al., 2020b); (3) agricultural water consumption, which is measured by effective irrigation area; (4) mechanical power, which is measured by the total mechanical power; and (5) pesticides and fertilizers, pesticide input is measured by the level of pesticide use, and fertilizer input is measured by the level of pure fertilizer use(Xu et al., 2021). All the input and output indictors are shown in Table 3.

Table 3 Variables and indicators description of agriculture eco-efficiency

As small-scale farming by individual owners is still a distinct feature of China’s rural economy, agricultural production is characterized by constant scale returns, and agricultural production has had negative effects on the environment. In this paper, the constant scale return model (CRS) with unexpected output is used to calculate AEE and to assign equal weight to both expected and unexpected outputs.

3.3 The panel smooth transition regression model specification

In the initial stage of agriculture and tourism integration, operators pay more attention to the direct economic benefits of agriculture than the sustainability of agriculture (Cui et al., 2020). They seldom consider the impact on the environment, and seldom take initiatives to reduce the use of chemical fertilizers, pesticides and other harmful environmental factors. Therefore, the effect of low integration level on reducing the input of harmful environmental factors may be not so significant. When the integration of agriculture and tourism enters the advanced stage, the ecological premium of agriculture is fully exploited and the ecological protection consciousness of operators is stimulated. Therefore, operators will consider adopting a more green and low-carbon way of production and operation. Therefore, the impact of agriculture and tourism integration on AEE will increase with the improvement of integration degree. In order to demonstrate the nonlinear impact of agro-tourism integration on the improvement of AEE, this paper adopts the Panel Smooth Transition Regression model (PSTR) (González et al., 2005). Core variables considered here include the degree of agricultural and tourism integration (ATL_it) and agricultural eco-efficiency (AEE_it). Also, we take into consideration the financial support for agriculture, industrialization levels and agricultural economic development levels factors as explanations of AEE (Hong et al., 2016; Wang et al. 2018). According to the existing literature, financial support for agriculture (WF_it) is measured by the ratio of local fiscal expenditures on agriculture, forestry and water conservation (Jiang, 2022). The industrialization level (ID_it) is measured by ratio of industrial added value to gross domestic product. The level of agricultural development (PIC_it) is measured by the per capita added value of agriculture (Jiang, 2022).Testing model is specified as follows:

$$\begin{aligned} InAEE_{{it}} = & \mu _{i} + \beta _{{01}} Inatl_{{it}} + \beta _{{02}} iNPIC_{{it}} + + \beta _{{03}} inID_{{it}} + \beta _{{04}} InWF_{{it}} \\ & + \left( {\beta _{{11}} InATL_{{it}} + \beta _{{12}} InPIC_{{it}} + \beta _{{13}} InID_{{it}} } \right)g\left( {q_{{it}} } \right)\left( {q_{{it}} ;\;r,\;c} \right) + \varepsilon _{{it}} \\ \end{aligned}$$

where i = 1…N, t = 1…T; N represents the number of sample provinces, and T represents the study period; μ_i represents individual fixed effects and ε_it is the error term (Tiba, 2019). g(q_it; r, c) is the transition function (Granger & Teräsvirta, 1993; Luukkonen et al., 1988), which is given as follows:

$$g\left( {q_{it} ;r,\;c} \right) = \left\{ {1 + \exp \left[ { - r\prod\limits_{j = 1}^{m} {\left( {q_{it} - c_{j} } \right)} } \right]} \right\},r > 0,\;c_{1} \le c_{2} \le \cdots \le c_{n}$$

(12)

In Eq. (12), q_it is a transition variable; r is the slope parameter, which determines the transition speed; c is the position parameter, which determines the threshold of parameter transition, the threshold value = EXP(c); and m is the number of position parameters (Chen & Chen, 2021). In the PSTR model, the coefficient of variable estimation is composed of linear β₀ and nonlinear β₁*g(•). When g(•) = 0, the model is in low regime. When g(•) = 1, the model is in a high regime. At the same time, when the value of the conversion function moves smoothly from [0, 1], the estimated coefficients are monotonically converted from β₀ to β₀ + β₁ which centers on c.

When the transition variable is the integration level ATL_it, the effect of ATL_it on AEE can be expressed as:

$$b_{ATL\;it} = \frac{{\partial InAEE_{it} }}{{\partial InAEE_{it} }} = \beta_{01} + \beta_{01} \times g\left( {q_{it} ;r,\;c} \right) + \beta_{11} InATL_{it} \frac{{\partial \left( {q_{it} ;r,\;c} \right)}}{{\partial \left( {InALT_{it} } \right)}}$$

(13)

When b_ATLit is greater than 0, with improvements in integration, AEE is constantly improved; therefore, integration level promotes the improvement of AEE. When it is negative, with improvements in integration, AEE declines continuously, therefore the level of integration inhibits the improvement of AEE.

Following research practice in this field (Gonzalez et al. 2017), a three-stage test is required before analysis. First of all, we take the Taylor expansion of g(q_it; r, c) where r = 0’ and then construct an auxiliary regression. Then we used the Wald, Fischer and likelihood ratio tests based on Colletaz and Hurlin’s insights (Anatolyev, 2004; Im et al., 2003). If the null hypothesis (H₀: r = 0) is rejected, there is a nonlinear relationship. The second stage involves testing the number of parameters. A null hypothesis test of r = 1 is first carried out, and if H₀ (r = 0) is rejected, a null hypothesis test of r = 2 is carried out until the null hypothesis r = r* cannot be rejected. Finally, the number of transition functions is obtained as r*. The third stage involves selecting the optimal model for estimation from models m = 1, r = r* and m = 2, r = r*.

3.4 Data source and study area

The integral development of agriculture and tourism generally occurs in areas abundant in agricultural and culture-historical resources (Ristić et al., 2019; Yang et al. 2021a). This paper focuses on 19 provinces in the two largest river basins in China, which are rich in agricultural and natural and cultural resources. 11 provinces in The Yangtze River Basin are used in this study, including Shanghai, Jiangsu, Zhejiang, Anhui, Jiangxi, Hubei, Hunan, Guizhou, Chongqing, Sichuan, Yunnan, while 8 provinces in the Yellow River including Shandong, Shanxi, Henan, Shaanxi, Inner Mongolia, Ningxia, Gansu, and Qinghai.

The research data used in this paper are drawn mainly from China’s statistical yearbook, tourism statistical yearbook, rural statistical yearbook, industrial statistical yearbook, and provincial statistical yearbooks and from the Statistical Bulletin of National Economic and Social Development (2008–2019). In addition, the National Bureau of Statistics, National Tourism Administration, and Ministry Of Agriculture and official provincial websites are used as sources of data supplementary. All data measured in monetary units have been deflated according to constant price levels in 2008.

3.5 Statistics description

The integration level of agriculture and tourism (ATL) is measured by the synergy degree model as shown above. Based on the synergetic degree model, ATL in each province is estimated from 2008 to 2019. Using the dataset and the Super-SBM model, AEE of each province is calculated from 2008 to 2019 at the same time. The basic average level of AEE and ATL in the study area over the study period is shown in Fig. 3. On the whole, AEE and ATL of 19 provinces have continuously improved over time, growing at an annual rate of 2.60% and 1.76%, respectively. Integration of agriculture and tourism is an important form of rural industries integration, which is beneficial for promoting rural economic growth and rural revitalization, so it is also strongly supported by the government. And in recent years, the central government in China has attached great importance to sustainable development of agricultural ecological environment; therefore, AEE has improved greatly, which can be seen in Table 4. The used variables in the PSTR model are displayed in Table 5.

Fig. 3

Average levels of AEE and ATL of the study area from 2008 to 2019

Table 4 AEE of each province in the study area from 2008 to 2019

Table 5 Description of variables in PSTR model

3.6 The unit roots testing

To avoid “pseudo-regression” or “pseudo-correlation” in the empirical results, the stationarity test of each variable is first carried out in this paper to ensure the unbiased and consistent empirical results. Unit root test is a common method to test stationarity of variables. In order to avoid the possible errors caused by a single test, three methods (IPS, LLC and ADF-Fisher) (Im et al., 2003) were adopted in this paper to conduct unit root test on data. The results of panel unit roots are shown in Table 6. From Table 6, we can see that all variables reject the null hypothesis of the existence of unit root, and the panel data have good stationarity.

Table 6 The Panel unit roots test

4 The empirical results and discussion

4.1 Nonlinearity analysis

4.1.1 Nonlinearity tests

Before PSTR model estimation, heterogeneity test, namely nonlinearity test, should be conducted to determine whether the model has nonlinear effect. If heterogeneity exists, it is more reasonable to use the PSTR model for estimation (Yao & Su, 2021); otherwise, it is considered that linear estimation should be carried out. In Table 7, model (1) is set for all 19 provinces, model (2) is set for 11 provinces in the Yangtze River Basin, and model (3) is set for 8 provinces in the Yellow River Basin. Using these 3 models, we can test whether ATL in different regions show differences in improvements in AEE. As can be seen from Table 7, for model (1), (2) and (3), when m = 1 and m = 2, the test values of LM_W, LM_F and LR all reject the null hypothesis so that the model is linear at the significance level of 1%.Therefore, it comes to a conclusion that agriculture and tourism integration has a nonlinear impact on agricultural eco-efficiency.

Table 7 Nonlinearity test of the PSTR

4.1.2 Residual nonlinearity test

Next, residual nonlinearity test is performed to determine the number of transformation functions r. The null hypothesis H₀ assumes that the PSTR model has only one threshold; the alternative hypothesis H₁ assumes that the PSTR model has two thresholds or more (Tiba, 2019). As can be seen from Table 8, none of the three models reject the null hypothesis, so they all have only one transformation function, that is r = 1.

Table 8 Setting test of the number r of conversion functions

After determining the number of transition functions r = 1, the number of position parameters should be determined as follows: in each model, the number of position parameters corresponding to minimum AIC and BIC values is the number of selected position parameters (González et al., 2005). According to this principle, it can be seen from Table 9, on the premise of r = 1, when m is set to the number 1, AIC and BIC values of the three models are all smaller than those when m is set to 2. Therefore, it can be concluded that the best mechanism combination of the three models is one transformation function and one threshold parameter (r = 1, m = 1).

Table 9 Test of the number m of threshold parameter

4.2 Empirical results

4.2.1 Results of PSTR

From the number of optimal transition functions and position parameters, the estimation results of the model using data for each province for the study period are as follows (Table 10):

• (1) For the whole study area, the linear part coefficient (β₀₁ = 0.0668) and the nonlinear part coefficient (β₁₁ = 0.4018) both are greater than zero, and significant at confidence levels of 1%, indicating that the agriculture and tourism integration can contribute to the improvement of AEE. When the conversion function g(ATL_it; r, c) = 0, the influence effect of agriculture tourism integration development was 0.0668(β₀₁), indicating that the model was in a low mechanism. When the function g(ATL_it; r,c) = 1, the total influence effect of agriculture tourism integration was 0.4686(β₀₁ + β₁₁), indicating that the model was in a high mechanism. Between low and high mechanisms, the impact effect of agriculture tourism integration is centered on the threshold value of 0.6821 (the threshold value = EXP(c), that is e^{−0. 3826}) of agriculture eco-efficiency, and smoothly transforms in the interval of [0.0668, 0.4686] with the change of its own state variables. This indicates that the integration of agriculture and tourism is not only conducive to the improvement of AEE, but also with the increase in integration level, its promoting effect on agricultural ecological efficiency presents an enhanced trend.

• 1. Combined with the analysis of AEE in the study area from 2008 to 2019(Fig. 3), the growth rate of AEE was relatively slow from 2008 to 2010, and the rate accelerated after 2011. From the perspective of agriculture and tourism integration history, although the No. 1 document of the Chinese Central Government proposed to develop rural tourism for the first time in 2007, at the beginning of the implementation of the policy, the integration of agriculture and tourism was mainly based on the simple model of “leisure agriculture” or “Nongjiale”, which was a simple and crude model of integration. Since then, leisure agriculture and rural tourism have played a great part in promoting the realization of multiple functions of agriculture and the integration of rural industries. Policies supporting industrial integration have been issued intensively since 2010, which have promoted the integral development of agriculture and tourism greatly, and the integration has played a more and more significant role in improving AEE. The influence coefficient transforms in the interval of [0.0668, 0.4686] during the study period. Meanwhile, we can see in Table 4, AEE of 2 provinces crossed the threshold value of 0.6821 in 2008, while AEE of 18 provinces crossed the threshold value in 2019. So it also proves the influencing mechanism of agriculture and tourism integration on improving AEE from an empirical perspective.

• 2. In the Yangtze River Basin, both the linear part coefficient (β₀₁ = 0.0712, P<0.01) and the nonlinear part coefficient (β₁₁ = 0.4832, P<0.01)) are greater than zero, indicating that the agriculture and tourism integration in this area can contribute to improving AEE. The linear part coefficient (β₀₁ = 0.0712) is higher than that of the whole research region 0.0668. The nonlinear part coefficient (β₁₁ = 0.4832) is also higher than that of the whole research region 0.4018. So we can say that the agriculture and tourism integration in the Yangtze River Basin has a greater improvement effect on AEE than the whole research region. The influence coefficient transforms in the interval of [β₀₁ = 0.0712, β₀₁ + β₁₁ = 0.5544] during the study period. When the integration level of provinces crossed the threshold value of 0.6879 (the threshold value = e^{−0. 3741}), we can see a continuous enhancement trend of the impact of the integration on AEE. The Yangtze River Basin is rich in water and forest resources with a high level of urbanization and economy. The relatively superior natural resources, climate conditions and relatively developed economy conditions promote people’s demand for agro-related leisure in this area. With the improvement of integration level, agricultural ecological value is further highlighted, which also promotes the enhancement of ecological consciousness of agricultural producers. So they will take initiatives to protect agricultural ecological environment as to achieve sustainable benefits.

• 3. As for the Yangtze River Basin, the linear part coefficient is 0.0002(not significant at all given confidence levels). The nonlinear part coefficient is 0.3022(significant at confidence level of 10%), which is lower than the nonlinear part coefficient of the whole research region 0.4018. The threshold of agricultural and tourism integration of provinces in the Yellow River Basin is higher than Yangtze River Basin and the whole study area(the threshold value e^{−0. 3281} = 0.7203). Only when the threshold value of 0.7203 is crossed, agriculture and tourism integration can improve AEE greatly. This may be due to the relatively weak economic foundation of most provinces in the Yellow River Basin, which makes it difficult to promote advanced agricultural technology. According to (Deng & Gibson, 2019), economy factor is important for the application of advanced agricultural technology. In addition, natural environment in this area is relatively harsh, and most provinces suffer from frequent droughts. In the Loess Plateau in the Yellow River Basin, soil erosion and sand storm damage has result in the eco-environmental deterioration. Because of the poor natural resources and climate conditions of this region, it is relatively difficult for agricultural tourism to be widely carried out. Therefore, in the early stage of integral development, the integration level of agriculture and tourism has no significant effect on agricultural ecological efficiency. And when the integration level crossed the threshold value of 0.7203, ecological capital of agriculture can create more value, which stimulates agricultural producers to pay more attention to agricultural sustainable development. Then, they will consciously reduce harmful environmental elements input in the producing process so as to improve AEE, and the promotion effect of agriculture tourism integration on AEE is enhanced.

• 4. From the perspective of control variables in Model (1), the impact of per capita agricultural added value on AEE in the whole study area is positive. And the coefficient of the linear part is 0.1183(P<0.01), and the coefficient of the nonlinear part is 0.1670(P<0.05)(β₀₂ = 0.1183, β₁₂ = 0.1670). The influence coefficient transforms in the interval of [β₀₂ = 0.1183, β₀₂ + β₁₂ = 0.2853] during the study period. This indicates that per capita agricultural added value contributes to improving AEE, and this influence is further enhanced with the increase in the level of agriculture tourism integration. As the impact of industrialization level on AEE, its coefficient changed from -0.0021 in the linear part to 0.2254 in the nonlinear part, but the linear part was not significant at all given confidence levels. It can be concluded that in the low level integration stage, industrial development may inhibit the improvement of AEE slightly, but with the increase in ATL, the promotion effect of industrialization level on AEE became more and more significant. The effect of financial support to agriculture on AEE was negative and significant in both linear and nonlinear parts (β₀₄ = -0.2017, β₁₄ = -0.1133); the parameter estimations of the linear part and nonlinear part are significant at confidence level of 5%. The influence coefficient transforms in the interval of [β₀₄ = -0.2017, β₀₄ + β₁₄ = -0.3150] during the study period, which indicated that financial support to agriculture inhibited the improvement of AEE, and this effect was further enhanced with the increase in ATL. To a large extent, China’s financial support for agriculture tends to subsidize petroleum agricultural elements such as pesticides, chemical fertilizers and agricultural machinery, which has a negative effect on AEE (Han & Zhang, 2020; Wang & Zhang, 2018).

Table 10 PSTR model estimation results

4.2.2 Results of the influencing mechanism test

In order to analyze the influence mechanism of agriculture and tourism integration on agricultural eco-efficiency, PSTR models (4) to (9) were constructed. The dependent variables of each model are crop sown area, effective irrigated area, agricultural employees, total power of machinery, fertilizer and pesticide usage. The independent variable is ATL, and results are shown in Table 11. Meanwhile, the control variables of above models are consistent with those of models (1) ~ (3). Data of the whole study area are used for testing the influencing mechanism test.

Table 11 PSTR model results of mechanism test

It can be seen from Table 11 that there is a nonlinear relationship between the integral level of agriculture and tourism and the input factors of AEE. Among them, crop sown area, the effective irrigated area and the agricultural employees all decrease with the deepening of the integration. The total mechanical power increased with the increase in the integration level at the beginning, but when the integration level crossed the threshold, the total mechanical power decreased with the increase in the integration level. This may be because when the integration level is low, the input of machinery replaces the input of agricultural labor force, which greatly reduces the input of agricultural labor force factors. However, the total power input of machinery shows an increase trend because the level of agricultural technology has not been greatly improved. With the improvement of integration level and agricultural technology level, the nonlinear part of total mechanical power input decreases after crossing the threshold value.

After analyzing the influence mechanism of ATL on the input factors of agricultural production, it is found that: Given the expected output, the reduction of input factors and unexpected output is limited at the initial stage of the integration, so the space for the improvement of AEE by the integration is limited. With the high level of integration, both input factors and non-expected output have a large space to reduce under the premise of expected output. Therefore, the promotion effect of agriculture and tourism integration on AEE shows a gradually expanding trend.

4.3 Endogeneity test results

The integration of agriculture and tourism can promote improvements in AEE, and conversely, improvements in AEE may also promote the agriculture and tourism integration. We used two methods to test the endogeneity of the model. First, we established a model with levels of agricultural and tourism integration lagging by two periods to demonstrate the effect of agricultural and tourism integration on improvements in AEE. It is assumed that only agricultural and tourism integration in the past can improve AEE in future. The results show that effects of agricultural and tourism integration on improvements in AEE are basically the same as those described above, indicating that the agriculture and tourism integration promotes improvements in AEE. Second, we use road network density as an instrumental variable for endogeneity analysis for two main reasons. On one hand, the government’s support for the integration of agriculture and tourism is reflected in the construction of transportation infrastructure. To integrate agriculture and tourism, quality transportation infrastructure must be built. Therefore, the integral level of agriculture and tourism can be represented by the density of highway networks to some extent. Second, the density of road networks does not directly affect AEE. The regression results from the first stage show that the statistical value of F is greater than 10 and passes a significance test of 5%, indicating that there is no weak instrumental variable problem. The integration level coefficient of the second stage is positive and significant, indicating that the integration promotes improvements in AEE.

5 Conclusion and implications

5.1 Conclusion

Based on provincial data of the Yangtze River basin and Yellow River basin in China from 2008–2019, this study applies PSTR model to test the impact of the agriculture and tourism integration on agricultural ecological efficiency, together with its impact mechanism. It can be seen that the impact of the integration of agriculture and tourism on AEE presents nonlinear characteristics with the integration level as the threshold. When its integration level was lower than the threshold value, it had a low promotion effect on agricultural eco-efficiency. However, when the integration level is high after crossing the threshold value, its promotion effect on AEE is very significant and shows an enhanced trend. And there are some differences of this impact in different regions. The agriculture tourism integration in provinces in the Yangtze River Basin has the most significant promotion effect on AEE. But the impact not is so an obvious in provinces in the Yellow River Basin when the level of agriculture and tourism integration is lower than the threshold value, only if the integration level cross the threshold value, it has a greater improvement effect on AEE. Based on the analysis of the influencing mechanism, it can be seen that the input of crop sown area, agricultural labor force, effective irrigation area and other factors decreases with the improvement of the integration level. Inputs such as pesticides and fertilizers did not decrease significantly in the initial stage of integration, but decreased rapidly after the level of agriculture and tourism integration crossing the threshold value. The total power of agricultural machinery rose in the initial stage of integration and fell rapidly after crossing the threshold value.

5.2 Theoretical implications

The theoretical enlightenments brought by this study are given as follows: On the one hand, this study expands the research content of influencing factors of agricultural eco-efficiency, indicating that we may explore the influencing factors from the perspective of industrial integration besides the input factors of agriculture; on the other hand, it also enriches the impact research of agricultural tourism integration, while most existing literature mainly concentrated on the effects of agri-tourism on the rural development and residents’ living environment (Ćirić et al., 2021; Lupi et al., 2017; Nooripoor et al., 2021). Meanwhile, it provides a new research perspective to reveal the inter-relationship between agriculture and tourism, while lots of studies are concentrated on the influences of agriculture on tourism. For example, Fleischer and Tchetchik (2002, 2005) have analyzed the impact of agriculture on tourism and found that farmers engaged in agri-tourism benefit from the existence of farm activities. Therefore, researches should pay attention on sustainable development of agriculture from the perspective of rural industry integration and multiple functions of agriculture. The integration of agriculture and other industries is an important way to expand the agricultural industrial chain, whick is also helpful to promoting sustainable development of agriculture.

5.3 Policy implications

The results of this study are of direct importance to policymakers in China, for the government plays a very significant role in promoting industries integration in rural areas (Ćirić et al., 2021; Su et al., 2019). As it is discussed above, the integral development of agriculture and tourism is helpful for improving agricultural ecological efficiency, which is also beneficial for achieving sustainable agricultural development and rural revitalization. Therefore, governments should further improve policies and institutional guarantee conditions for the development of agriculture and tourism integration. Considering the particularity of agricultural development, the government should increase support and play a leading and coordinating role in the integration process, but also coordinate the interests of enterprises, farmers and other stakeholders. Second, local governments at all levels should formulate relevant policies based on local conditions in view of regional differences in the improvement effect of the integration on agricultural eco-efficiency, so as to better play the positive role of agro-tourism integration in improving agricultural eco-efficiency. Agriculture and tourism resources are abundant in provinces in the Yangtze River basin, the agriculture and tourism integration has good development foundation, so the government should further perfect policy support of industries integration development, strengthen the regulatory functions of integration development, coordinate the relationship between different stakeholders. Due to the geographical, ecological environment, economic conditions of provinces in the Yellow River Basin, the foundation of agriculture tourism integration is relatively weak. Therefore, the government departments can promote the integration by increasing financial support to give farmers who engaged in agri-tourism business, establishing some influential agri-tourism brands, and strengthening the agricultural ecological education and ecological management in agriculture, as research found that governance and education are important on improving AEE (Bayyurt & Yılmaz, 2012). At the same time, it should stimulate agricultural leisure consumption through marketing propaganda, so as to raise agricultural output and promote rural economic and environmental development.