Farmers’ Endowments, Technology Perception and Green Production Technology Adoption Behavior

Abstract

The motivation of this paper is to explore the influence of farmers’ endowments and technology perceptions on farmers’ green production technology adoption behaviors. Based on a survey of 471 grain farmers in the main grain-producing areas of Shandong Province, the Heckman model was used to empirically analyze the influence of farmers’ endowments on their adoption behavior of green production technologies, and the moderating effect of technology perception on the relationship between the two was further investigated. The study showed that: (1) More than half of the farmers adopted more than three green production technologies, and only 11.5% did not adopt them. In terms of adoption structure, the adoption rate of green pest control technology was the highest at 67.7%, while the adoption rate of moderate deep pine technology was the lowest at 32.7%. The adoption structure should be further optimized. (2) After controlling for sample selection bias, farmers’ endowments have a significant positive effect on farmers’ green production technology adoption behaviors. The ordinary least square regression overestimates the main effect of farmers’ endowments by 8.5% compared with the Heckman model. (3) Technology cognition can effectively promote the positive effect of farmers’ endowments on green production technology adoption, in which the moderating effect of technology–economic cognition is higher than that of technology–environmental cognition. (4) There is heterogeneity in the effects of farmers’ endowments and technology cognition on different adoption subjects and technology types. The above findings provide an in-depth explanation for activating the endogenous drivers of green agriculture development.

1. Introduction

China’s agricultural development has made world-renowned achievements in recent years, basically realizing the transformation from absolute shortage to matching supply and demand, as well as from purely pursuing quantitative growth to considering the quality. However, agricultural pollution and high yields go hand in hand. The traditional crude agricultural production methods have caused agricultural surface water pollution and have seriously restricted the sustainable development of agriculture [1]. Agricultural development should not only eliminate the factors in the ecological environment that create new debts, but also gradually pay off old debts in order to fight the battle of agricultural surface pollution prevention and control. With the increasing constraints of arable land, fresh water, and other resources, it is an essential focal point to achieve high-quality agricultural development. There is an urgent need to optimize the agricultural green development technology system for efficiency. Comprehensive promotion of agricultural waste reduction, control of agricultural surface pollution, focus on soil and water conservation, and protection of the environment, as the characteristics of green production technology, have become necessary for the effective management of agricultural pollution [2]. Agricultural green production technology refers to using environmentally friendly, efficient, and sustainable techniques and methods in agricultural production, in order to minimize the negative impact on the environment and protect the ecosystem to the greatest extent possible. These techniques include cultivation methods, fertilization, land fragmentation, and crop rotation technologies.

Farmers’ adoption of green production technology is an inevitable requirement for the cultivation and further and strengthening of the new momentum of green agricultural development. Currently, China’s agricultural pollution management is still overly dependent on the government, and the lack of a micro-governance mechanism has become an outstanding shortcoming [3]. Farmers are the main body of agricultural technology application, and standardizing and guiding the production-oriented environmental behavior of farmers and improving the adoption rate of farmers are the keys to effective promotion of green production technology [4,5]. With the deepening of rural reform and the gradual improvement of rural markets, driven by both industrialization and urbanization, farmers have begun to evolve from homogeneity to heterogeneity. Different types of farmers have different endowments and differences in their cognition of agricultural technologies [6]. Identifying the influence of farmers’ endowments on the adoption behavior of green production technologies and their mechanism of action becomes the key to further improving the demand-oriented agricultural technology promotion mechanism.

Farmers’ production behaviors are important factors affecting the level of sustainable development. The relationship between farmer endowment and farmer technology adoption has been explained and verified by studies [7,8,9]. Farmer endowment is an essential factor in farmer technology adoption behavior. Regarding human capital endowment, age and household size are positively associated with adopting new technologies [10]. Farmers of higher ages have rich knowledge and experience in the field, and those with a larger household sizes break the labor constraints of new technology adoption [10]. Moreover, the off-farm employment behavior of family members can inhibit farmers’ ecological production behaviors. Regarding household endowment, farmers’ livelihood patterns influence their production technology perceptions [7], and farmers with a larger scale of operation adopt new technologies to a higher extent [8]. As the scale of cultivation expands, farmers may face higher initial input costs for adopting new technologies. However, they can obtain greater marginal returns to offset the costs and achieve economies of scale [9]. Regarding social capital endowment, farmers’ technology selection behavior has a clear tendency to cluster, which is an apparent herd behavior. Interactions among farmers with different knowledge backgrounds can strengthen their recognition of specific technologies, so farmers in social networks are more likely to adopt new technologies [11]. Farmer associations or cooperatives can reduce farmers’ risk perceptions of new technologies, and thus increase their willingness to adopt them [12]. Social capital spreads the risk of agricultural technology and influences farmers’ agricultural technology adoption by increasing the availability of physical capital, human capital, and market information. Scholars have explored the influence of farmers’ technology perceptions on adoption behavior from different perspectives and by using different methods [13,14]. Technology perceptions reflect farmers’ understanding of agricultural production technologies’ adoption methods and effects. These can be subdivided into convenience perceptions, technology effect perceptions, and technology risk perceptions [13]. A correct and comprehensive technology perception positively impacts farmers’ adoption behaviors, and the higher the level of a farmer’s knowledge, the higher the probability of adoption [14]. From the perspective of research methods, most of the existing research results treat farmers’ adoption behaviors of agricultural technology as a discrete choice problem. Probability models such as logistic, probit, and tobit, are used to investigate the significant factors influencing whether farmers adopt green production technology. Logical explanations are provided regarding the influencing factors’ magnitudes, directions, and significances. However, such probability models are conditionally applicable, and some green production technologies are a “collection” of technologies. In this case, farmers’ adoption behaviors are generally divided into “adoption” and “non-adoption”, and the more profound behavioral logic is not explored. In recent years, some scholars have further explored the degree of adoption (quantity or density) based on the study of farmers’ adoption or non-adoption, and this exploration gradually deepened from shallow description to deeper logical exploration. In summary, the relationship between farmers’ endowment and technology adoption, technology cognition and technology adoption has been demonstrated and discussed, laying the theoretical foundation and inspiring research ideas for this study. Does the promotion effect of farmers’ endowment on technology adoption also apply to green production technology? Can technology perceptions moderate the effect of farmers’ endowments on adopting green production technologies? Studies have yet to explore these questions.

Accordingly, in order to explore how a farmer’s endowment affects their green production technology adoption behavior, this paper constructs a mathematical model of farmers’ green production technology adoption and proposes a research hypothesis. Based on the survey data of 471 grain farmers in the central grain-producing regions of Shandong Province, the Heckman model is used to empirically analyze the influence of farmers’ endowments on farmers’ green production technology adoption behaviors, and it further investigates the moderating effect of technology cognition on the relationship between the two. Compared with existing studies, the marginal contributions of this paper are as follows: first, the inclusion of technology cognition in the analytical framework of the influence of farmers’ endowments on green production technology adoption fully considers the moderating effect of farmers’ technology cognition and enriches the existing studies on farmers’ green production technology adoption. Second, since whether farmers are selected for the sample is not an exogenous, purely random event, there is a problem of sample self-selection. Heckman’s sample selection model can correct the estimation bias caused by this bias [15]. Therefore, the Heckman model is used for parameter estimation, which effectively avoids the problem of sample self-selection in farmer surveys and enhances the reliability of the study findings through robustness and endogeneity tests. Exploring the relationship between farmers’ endowments, technology perceptions, and adoption behaviors can provide a reference for effectively solving the problem of the disconnect between green production technology supply and farmers’ demand, as well as for controlling pollution from the source of agricultural production.

The remainder of the paper is structured as follows. Section 2 contains theoretical analysis, Section 3 contains research methodology, Section 4 contains empirical results and analysis, Section 5 contains discussion and policy insights, and Section 6 contains the conclusions.

2. Theoretical Analysis

Farmer endowment refers to the resources and capabilities possessed by family members and the whole household, such as household head characteristics, family characteristics, social capital, economic status, and economic capital endowment of land, which contribute to the attitudes and capabilities of farmers to adopt green production technologies. Drawing on the criteria for classifying farmer endowment, and combining the characteristics of farmer green production technology, farmer endowment is classified into the human capital endowment, economic capital endowment, and social capital endowment. The human capital endowment mainly includes the gender, age, education, farming experience, and health status of the household head. The head of the household, as the spokesperson and representative of the farmer, can exercise the right to make agricultural production and management decisions on behalf of the farmer, and often determines the willingness and ability of the farmer to adopt new technologies [16]. Male, younger, and more educated household heads tend to accept the risks associated with replacing traditional technologies with new technologies and show positive acceptance of new technology diffusion. However, some studies suggest that women may have a more positive attitude toward new technologies than men [17]. In practice, the adoption of green production technologies means giving up traditional production methods. Some environmentally friendly agricultural technologies deviate from traditional agricultural production experience and have a long return on investment cycle, even affecting the level of farmers’ incomes, so that older farmers with longer farming experience are prone to technology-dependent inertia. On the contrary, younger and more educated household heads have a stronger risk tolerance and show a positive attitude toward promoting green production technologies.

Economic capital endowment refers to the factors of production owned by or at the disposal of farmers, which have a direct impact on their production and management decisions, including plot fragmentation, total arable land area, soil fertility, mechanization, and irrigation accessibility. Specifically, with the expansion of farming area, farmers are more inclined to adopt new technologies as the economies of scale become more significant and the marginal rewards of adopting new technologies are greater. The higher the fragmentation of plots and the more tedious the labor procedures, the less likely farmers are to adopt new technologies. Farmers with fertile soils and easily irrigated land tend to have higher productivity, and “yield-enhancing” agricultural technologies are not attractive, so farmers with fertile and easily irrigated land are less motivated to adopt new technologies. Farmers with large agricultural machinery (seeders and harvesters) are more likely to adopt labor-saving technologies and are more receptive to new technologies such as conservation tillage.

Social capital endowment is an intangible expression of capital that exists in social and geographical networks, and it can improve the efficiency and integration of society through interpersonal cooperation [15]. Chinese rural society is based on long-term “relationships” among farmers, who naturally form their own social networks. In this study, three indicators are used to characterize the social capital of rural households: position, frequency of communication, and participation in associations or cooperatives. Specifically, farmers who have family members who are village cadres or agricultural technicians tend to have better information about new agricultural technologies, are more likely to adopt new technologies, and play the role of technology disseminators in their social networks. Social networks have the function of disseminating information and playing a learning role. Communication between farmers and their neighbors and friends can facilitate the diffusion of new technologies, reduce the perception of risk of new technologies, and thus increase the willingness to adopt them. As the promotion of green production technologies increases, farmers who communicate frequently with their neighbors or friends tend to have a higher degree of knowledge about new technologies and are more likely to adopt them. Agricultural associations and cooperatives pay more attention to new agricultural technologies and accept green production technologies earlier, providing a platform to support small-scale farmers to connect with modern agriculture. Farmers who join agricultural associations or cooperatives have more opportunities to contact modern machinery dedicated to green production technologies through technical exchanges and other means, and they have more opportunities to learn and understand new technologies; thus, they are more likely to adopt green production technologies.

Technology cognition is the extent to which farmers know about the adoption method and effects of a green production technology [7,18]. Farmers cannot predict the potential benefits and additional costs of adopting a green production technology, but can only acquire certain technology perceptions from the external environment, such as through neighborhood demonstration, including technology economic perceptions and technology environmental perceptions. The relationship between technology perceptions and technology adoption by farmers’ endowments has been verified by existing studies [19]. Economic cognition includes convenience cognition, operational effectiveness cognition, and income generation cognition. Generally speaking, the easier it is to obtain and use the machinery and equipment, the higher the farmers’ evaluations of the operational effects, and the stronger the perception of income generation, the greater the possibility of farmers adopting green production technology [20]. It can be inferred that the economic perception of technology can effectively promote a positive relationship between farmers’ endowments and the adoption of environmentally friendly technologies. The environmental perceptions of technologies can be divided into soil conservation perceptions, pollution reduction perceptions, and ecological protection perceptions. Green production technology can maintain soil and water, slow down the trend of ecological deterioration, and reduce the environmental pollution problems caused by straw burning, which has obvious ecological and environmental benefits. The higher a farmer’s perception of the ecological and environmental benefits of green production technology, the more likely they are to adopt the technology. Thus, it can be seen that the perception of the environmental nature of the technology reinforces the positive influence of farmers’ endowments and the adoption of environmentally friendly technologies.

Combined with the theoretical analysis of the influence mechanism, this paper proposes two research hypotheses H1: Farmers’ endowments positively influences farmers’ green production technology adoptions. H2: Technology perception plays a moderating role in farmers’ endowments influencing green production technology adoption.

3. Research Methodology

3.1. Research Area and Data Sources

(1) Research area: Shandong is an important grain producing area in the northern region of China, with 6% of the country’s arable land and 1% of its freshwater resources, contributing 8% of grain production. Its wheat and corn production has outstanding advantages in the country. The Department of Agriculture and Rural Affairs issued the “Guiding Opinions on the Green, High Quality, and Efficient Creation of Grain in 2019”, focusing on the gateway area, the demonstration area, and the radiation area, and which is committed to the creation of a typical model of green grain production. Therefore, the sample of wheat and corn growers in the main grain producing areas of Shandong Province is representative.

The research data were mainly obtained from the thematic research activities conducted by the research team in October–November 2020 in the main grain producing areas of Shandong Province. Based on the consideration of geographical location and farmer differentiation, the sample selection was mainly oriented toward wheat and corn growers in green production technology promotion demonstration areas: Pingyuan, Ningjin, and Qingyun counties in Dezhou, Hanting, Gaomi, and Xiashan districts in Weifang, and Haiyang and Zhaoyuan counties in Yantai. The map of the research area is shown in Figure 1.

(2) Data sources: Before conducting the large-scale survey, 50 households of grain farmers were first selected for a pre-survey, and the questionnaire was adjusted and improved in response to the survey returns. In order to distribute the sample evenly in the overall population, this study adopted a sampling method combining stratified sampling and simple random sampling. Among them, stratified sampling was conducted based on the number of farmers in the village. The number of questionnaires distributed was determined based on the number of farmers in each village. Firstly, according to the principle of stratified sampling, two to four townships were selected in each sample county, and two to three villages were selected in each township. Second, we selected multiple samples in each village based on the principle of random sampling. The above samples were selected to have good representativeness for studying the adoption of green production technology among farmers in Shandong province. In order to improve the quality of the questionnaire survey, the survey was conducted in the form of face-to-face question and answer for the sample food growers. Members of the subject team were responsible for asking questions and recording the answers. Questionnaires that were completed entirely and did not have logical gaps were retained, and those that were incomplete and had logical problems were eliminated. A total of 500 questionnaires were distributed, of which 471 were valid, with a valid questionnaire rate of 94.2%. The descriptive statistics of the data are shown in

Table 1. Basic characteristics of the sample farmers.

Features		Sample Size (Households)	Proportion (%)
Farmer type	Small-scale pure farmers	135	28.7%
	Part-time business	227	48.2%
	Large professional	109	23.1%
Association or cooperative join	Yes	225	47.8%
	No	246	52.2%
Farming experience (years)	[0,10)	57	12.1%
	[10,30)	255	54.1%
	Over 30 years	159	33.8%
Types of green production technologies	Pest and disease green control	319	67.7%
	Soil testing and fertilization	202	42.9%
	No-till seeding with less tillage	244	51.8%
	Moderate deep relaxation	154	32.7%
	Straw return to the field	307	65.2%
Arable land area used for food production (mu)	[0,10)	178	37.8%
	[10,50)	184	39.1%
	More than 50 mu	109	23.1%
Are there village officials	Yes	71	15.1%
	No	400	84.9%
Number of operating land blocks (blocks)	[0,5)	307	65.2%
	[5,10)	157	33.3%
	More than 10 pieces	7	1.5%
Number of green production technology adoption (items)	0	54	11.5%
	1	60	12.7%
	2	100	21.2%
	3~4	225	47.8%
	5	32	6.8%

.

Drawing on the analysis of the farmer differentiation [21], farmers were classified into pure farmers, part-time farmers, part-time non-farmers, and non-farmers based on differences in household income structure. Non-farmers have left agricultural production, so non-farmers were not included in the sample. In the survey sample, part-time farmers accounted for 48.2%, indicating that part-time food production was still dominant. A total of 76.9% of the farmers had a business area of less than 50 mu (The conversion units of mu and ha are as follows: 1 ha = 15 mu) and had not yet realized large-scale food production. A total of 15.1% of the farmers have village cadres among their family members, and 47.8% of the farmers have joined agricultural associations or cooperatives to connect with modern agriculture. In terms of farming experience, 33.8% of the heads of households have more than 30 years of farming experience and have rich practical experience. A total of 65.2% of the farmers operate less than 5 plots of arable land, and the degree of fragmentation of arable land is not high, indicating that the sample area has basically realized contiguous operation in food production. In terms of the number of adoption, more than half of the farmers adopted more than three kinds of green production technologies, while 11.5% of the farmers did not adopt them. In terms of adoption structure, the adoption rate of green pest prevention and control technology, minimum tillage and no-tillage seeding technology, and straw return technology is more than half. The adoption rate of green pest prevention and control technology is the highest at 67.7% and the adoption rate of moderate deep pine technology is the lowest at 32.7%, indicating that there is a structural imbalance in the adoption of green production technology by farmers which needs to be further optimized.

3.2. Model Setup

The Heckman two-stage model is suitable for solving endogeneity problems caused by sample selection bias [22]. Since whether farmers are selected into the sample is not an exogenous, purely random event, there is a problem of sample self-selection. Heckman’s sample selection model can correct for this estimation bias [23]; therefore, this sample selection model was used.

The first stage of the model is a probit model including the full sample, and the selection equation is constructed to estimate the probability of farmers adopting or not adopting green production technology. Based on this, the inverse Mills ratio, which corrects for the selection bias of the sample, is calculated based on the selection equation. The selection equation in the first stage can be expressed as Equation (1). In Equation (1), $probit$ indicates the probability of adoption of green production technologies by farmers, $gift$ is the farmer’s endowment, $cognition$ is the technology perception. $\alpha ,\beta ,\gamma$ is the coefficient to be estimated, $Z$ is the vector of control variables, and $\epsilon$ is the random disturbance term.

$$probit(use=1)=\alpha \times gift+\beta \times cognition+\gamma \times Z+\epsilon$$

(1)

Referring to relevant study [24], in the sample selection, the effect of farmers’ endowments and technology perceptions on the number of green production technologies adopted was estimated using a maximum likelihood estimation method with the inverse Mills ratio. Farmers’ endowments, technology perceptions, and the control variables were calculated from the selection equation as explanatory variables, and the number of green production technologies adopted was the explanatory variable. The expressions of the resulting equation are shown in Equation (2). In Equation (2), $adoption$ indicates the number of farmers adopting green production technologies, $gift$ is the farmer’s endowment, $cognition$ is the technology perception, $gift\times cognition$ is the interaction term, and $\lambda$ is the inverse Mills ratio.$\alpha ,\beta ,\gamma ,\delta$ is the coefficient to be estimated, and $\omega$ is the random disturbance term.

$$adoption=\alpha \times gift+\beta \times cognition+\gamma \times gift\times cognition+\delta \times \lambda +\omega$$

(2)

3.3. Variable Selection

Dependent variable: based on the definition and classification of green production technologies by the United Nations Environment Programme, and with full consideration of the production characteristics of food crops in the sample areas, green pest prevention, control technologies, and soil testing and fertilizer application technologies were selected from the three production links of pest and disease, drug application, and fertilization and soil management, respectively. We also subdivided the conservation tillage technologies into three categories: no-till sowing with less tillage, moderate deep loosening, and straw return. Based on the Heckman model, the dependent variables were “whether to adopt” and “degree of adoption”. Among them, “whether to adopt” is a dummy variable, where 0 means that the farmer did not adopt any green production technology, and 1 means that the farmer adopted 1 or more of the 5 types of technologies: green pest control technology, soil testing and fertilizer application technology, less tillage and no-tillage seeding, moderate deep loosening, and straw returning to the fields. The “degree of adoption” was measured by the actual number of farmers’ adoptions, and the values were assigned from 1 to 5 in order.

Core independent variables: (1) Farmer endowment contains three variables: human capital endowment, economic capital endowment, and social capital endowment. Combined with the above theoretical analysis, the five variables of the gender of household head, age of household head, education level, farming experience, and health status were used to characterize the human capital endowment of farmers. The values are derived by extracting the common factor from each question item. The five variables of plot fragmentation, total arable land area, soil fertility, mechanization, and irrigation convenience were used to characterize the economic capital endowment of farmers. The values are derived by extracting the common factor from each question item. Drawing on the research results [15,25,26,27], the social capital endowment of farmers was characterized by position in the village, frequency of technical exchange, and cooperative participation. The values are derived by extracting the common factor from each question item. (2) Technology perception. Based on the study of the classification of technology cognition [28,29,30] and the basic characteristics of green production technology, technology cognition was subdivided into technology–economic cognition and technology–environmental cognition. Technical–economic cognition includes convenience cognition, operational effect cognition, and income generation cognition, while technical–environmental cognition includes soil protection cognition, pollution reduction cognition, and ecological protection cognition.

Control variables: (1) Land transfer. Farmers’ land transfer behaviors will inhibit the adoption of green production technologies, as measured by “whether to transfer in or out of the land” [31]. If farmers transfer out of their land, they may no longer or rarely operate their own farming business, so they will not adopt environmentally friendly agricultural technologies with high conversion costs; if farmers transfer into their land, they are more likely to adopt “short-sighted” behavior in their transferred land due to the current poorly regulated transfer and short transfer cycle. (2) Technical training. In this study, “whether they have participated in training on green production technology” was used to reflect farmers’ technical training. Technical training can improve farmers’ understanding of green production technologies and reduce their perceived risk [32], so farmers are more likely to adopt green production technologies. (3) Technology promotion. Whether farmers have received technical extension activities by professional and technical personnel is reflected by “whether they have received green production technology extension activities”. Agricultural technology extension personnel directly serve farmers and have rich experience and professionalism in agricultural production; they can personally disseminate agricultural technology information to farmers by providing technical guidance, consultation, demonstration, etc. Therefore, farmers who have received a promotion from professional technicians are more likely to adopt new technologies. (4) Government subsidies. In this study, the variable of government subsidies was measured by “whether the local government provides green production technology subsidies”. Technology subsidies can reduce the replacement cost of adopting new technologies and increase the expected benefits for farmers, which positively affects their adoption behavior. The specific definition and assignment of each variable are shown in

Table 2. Variable descriptions and descriptive statistics.

Variables	Variable Name	Variable Description and Assignment (Unit)	Average Value	Standard Deviation
Dependent variable	Whether	Whether to adopt green technology, No = 0, Yes = 1	-	-
	Degree	Number of green production technologies adopted (items)	2.66	1.46
Human capital endowment	Gender	Gender of farmer, Female = 0, Male = 1	-	-
	Age	Age of farmer (years)	50.54	8.53
	Education	Elementary school and below = 1, junior high school = 2, high school = 3, high school above = 4	1.77	0.74
	Years	Farming years of farmer (years)	25.80	12.87
	Health	Health status of farmer, Unhealthy = 0, Healthy = 1	-	-
Economic capital endowment	Plot dispersion	Number of plots (blocks)	4.07	2.04
	Land area	Total area of cultivated land (mu)	24.19	24.27
	Soil fertility	Barren = 0, Fair = 1, Fertile = 2	1.65	0.52
	Mechanization	No = 0, Yes = 1	-	-
	Irrigation	Irrigation convenience, Likert five-point scale	3.31	0.96
Social capital endowment	Position	Position in the village, No = 0, Yes = 1	-	-
	Communication	Technical communication, No = 0, Yes = 1	-	-
	Cooperative	Cooperative participation, No = 0, Yes = 1	-	-
Technology economic perception	Convenience	Convenience perception, Likert scale	3.38	0.96
	Effectiveness	Assignment Effectiveness Perception, Likert scale	3.39	0.94
	Income	Income-generating perception, Likert scale	3.35	0.94
Technology environmental perception	Protecting soil	Protecting soil perception, Likert scale	3.56	0.91
	Pollution reducing	Pollution reduction perception, Likert scale	3.56	0.99
	Ecological	Ecological conservation perception, Likert scale	3.37	0.96
Control variables	Transfer	Land transfer, No = 0, Yes = 1	-	-
	Training	Technical training, No = 0, Yes = 1	-	-
	Promotion	Technology promotion, No = 0, Yes = 1	-	-
	Subsidy	Government subsidy, No = 0, Yes = 1	-	-

.

4. Empirical Results and Analysis

4.1. Estimation Results and Analysis of Main Effects of Farmer Endowments

In order to investigate the influence of farmers’ endowments on their adoption and the degree of adoption, this paper used Stata14.0 software to analyze the sample data and compared the results of the ordinary least square regression (OLS for short) model with those of Heckman model. The results are shown in

Table 3. Influence of farmers’ endowments on farmers’ green production technology adoption behaviors.

Variables	OLS Regression Model		Heckman Model 1		Heckman Model 2
	Degree	Degree	Whether	Degree	Whether	Degree
Farmers’ endowments	0.552 ***		0.316 ***	0.468 ***
Human capital endowment		0.241 ***			0.169 **	0.199 ***
Economic capital endowment		0.405 ***			0.112	0.423 ***
Social capital endowment		0.053			0.154	−0.017
Transfer	−0.069	0.001	−0.109	−0.082	−0.110	−0.046
Training	0.216	0.267 **	0.591 ***	0.065	0.597 ***	0.023
Promotion	−0.335 **	−0.326 **	−0.317	−0.232 *	−0.315	−0.241 *
Subsidy	0.606 ***	0.646 ***	0.911 ***	−0.026	0.916 ***	0.096
Constant term	2.252 ***	2.150 ***	0.772 ***	3.134 ***	0.765 ***	2.901 ***
R2	0.215	0.219
Inverse Millsby			0.084 **		0.063 **
Wald test			53.510 ***		76.020 ***

Note: *, **, and *** indicate significant at the statistical level of 10%, 5%, and 1%, respectively.

.

The results in Table 3 show that the effects of farmers’ endowments on green production technology adoption behaviors in the OLS and Heckman models are 0.552 and 0.468, respectively, and both are significant at the 1% level, indicating that there is a positive effect of farmers’ endowments on their green production technology adoption, and H1 is verified. In the Heckman model, the inverse Mills ratio (IMR) is significant at the 5% level. The main effect of OLS on farmers’ endowments was overestimated by 8.45%, and the Heckman model effectively corrected for the sample selectivity bias of farmers’ green production technology adoption behaviors. Further, based on the previous analysis, farmer endowments were subdivided into human capital endowments, economic capital endowments, and social capital endowments, and their effects on farmers’ adoption behaviors of green production technologies were investigated separately. The estimation results of Heckman Model 2 show that both the human capital endowment and the economic capital endowment of farmers significantly and positively affect the adoption behavior of green production technology at the 1% level, i.e., the higher the level of human capital endowment and economic capital endowment, the more farmers tend to adopt green production technology. The effect of social capital endowment on the adoption of green production technologies by farmers was not significant.

4.2. Analysis of the Moderating Effect of Technology Perception

To investigate whether technology cognition can effectively moderate the effect of farmers’ endowments on green production technology adoption, this paper adds the cross-product term of farmers’ endowments and technology cognition to the regression model of farmers’ endowments, technology cognition, and green production technology adoption behaviors. The Heckman model was then used to investigate the magnitude and direction of the moderating effect of technology cognition. Model 3 shows the mediating effect of technology cognition, Model 4 shows the mediating effect of technology economic cognition, and Model 5 shows the mediating effect of technology environment cognition. The estimated results are shown in

Table 4. Estimated results of the moderating effect of technology perception.

Variables	Heckman Model 3		Heckman Model 4		Heckman Model 5
	Whether	Degree	Whether	Degree	Whether	Degree
Endowment	0.322 **	0.424 ***	0.346 ***	0.433 ***	0.311 **	0.435 ***
cognition	−0.022	0.136 ***
Endowment × cognition	0.006	0.066 *
economic			−0.111	0.142 ***
Endowment × economic			−0.073	0.049 *
environment					0.066	0.106 **
Endowment × environment					0.080	0.081 *
Transfer	−0.105	−0.001	−0.113	−0.016	−0.094	0.005
Training	0.596 ***	−0.033	0.625 ***	−0.021	0.589 ***	−0.036
Promotion	−0.320	−0.237 *	−0.343	−0.234 *	−0.308	−0.239 *
Subsidy	0.923 ***	0.024	0.945 ***	−0.019	0.906 ***	0.048
Constant	0.759 ***	2.982 ***	0.780 ***	2.993 ***	0.754 ***	2.956 ***
Inverse Millsby	0.075 **		0.076 **		0.077 **
Wald test	64.000 ***		64.180 ***		61.620 ***

Note: *, **, and *** indicate significant at the statistical level of 10%, 5%, and 1%, respectively.

.

From Model 3 in Table 4, the regression coefficient of technology awareness is 0.136, which is significant at the 1% level, and the regression coefficient of the cross-product term between farmer endowment and technology awareness is 0.066, which is significant at the 10% level. It indicates that technology awareness can effectively promote the positive relationship between farmer endowment and environmentally friendly technology adoption, and H2 is verified. Meanwhile, the coefficient of the cross-product term is small, indicating that the moderating effect of technology cognition is weak. Farmers’ cognition of green production technologies is a gradual process. At the early stage of green production technology promotion, farmers’ cognition of green production technology is not sufficient due to the limitation of knowledge level, which indicates that green production technology promotion is long-term, and agricultural extension departments need to be prepared for a long-lasting effort.

Technological cognition contains techno–economic cognition and techno–environmental cognition. The estimation results of Model 4 and Model 5 in Table 4 show that the regression coefficient of techno–economic cognition (0.142) is more significant than that of techno–environmental cognition (0.106). They are significant at 1% and 5% levels, respectively, indicating that the moderating effect of technological–economic cognition is stronger. Green production technology promotion should include a good cost–benefit analysis for farmers and focus on popularizing the economics of technology. The regression coefficient of the cross-product of farmer endowment and technology–economy cognition (0.049) is smaller than that of the cross-product of farmer endowment and technology–environment cognition (0.081), and both are significant at the 10% level. This indicates that the moderating effect of environmental perceptions is higher than for economic perceptions. In the face of the current increasing proliferation of pollution in agricultural production, farmers are increasingly concerned about the environmental benefits of production technologies.

4.3. Heterogeneity Analysis

(1) A heterogeneity analysis of farmers’ adoption behaviors by differentiating adoption subjects was performed. Most existing studies on the adoption behavior of green production technologies are aimed at retail households [13,14]. Few studies have focused on the adoption behavior of new agricultural management subjects such as large professional households and family farms. Under the influence of both technological development and educational modernization, farmers’ overall quality will be significantly improved, and their cognition of new technologies will be significantly changed. This paper discards the traditional “one-size-fits-all” research perspective and further explores the differences in the adoption behavior of different types of farmers in the context of farmer differentiation. The sample farmers are divided into small-scale farmers, part-time farmers, and large professional farmers based on the scale of their farming operations and the proportion of their farming income to their household income (According to the Shandong Province "on further standardizing the implementation of subsidies for large grain farmers" (Lu agricultural finance word [2016] No. 53), the region’s large grain farmers, with regard to wheat or rice planting area, must plant 50 mu or more as a condition for identification) in order to explore the heterogeneity of their adoption behaviors. The heterogeneous farmers’ demand characteristics were used to explore the low adoption rate of green production technologies and the disconnect between agricultural technology supply and farmers’ demand at this stage. The estimated results of the Heckman model for the regression of farmer heterogeneity grouping are shown in

Table 5. Estimated results of farmer heterogeneity.

Variables	Model 6 Small-Scale Farmers		Model 7 Part-Time Farmers		Model 8 Large Professional Farmers
	Degree	Whether	Degree	Whether	Degree	Whether
Endowment	−0.014	−0.091	0.287 ***	0.435 **	−0.250	−3.044 *
cognition	0.002	−0.148	0.264 ***	0.041	−0.058	−0.931
Endowment × cognition	0.021	−0.238	0.203 *	0.055	0.130	1.072
Transfer	0.057	−0.094	0.186	−0.026	−0.195	−0.928
Training	−0.269	1.194 ***	0.329 *	0.384	−0.112	2.057 **
Promotion	−0.022	−0.560 *	−0.448 **	−0.415	0.160	8.846
Subsidy	0.278	0.891 ***	−0.261	0.882 ***	0.007	1.479 ***
Constant	2.088 ***	0.189	3.107 ***	1.133 ***	3.955 ***	1.911
Observations	135		227		109
Inverse Millsby	0.055		0.044 *		−0.192
Wald test	3.030		34.310 ***		4.620

Note: *, **, and *** denote significant at the 10%, 5%, and 1% statistical levels, respectively.

.

As can be seen from Table 5, the variables of farmer’s endowment and technology cognition in Model 6 are not significant. The original hypothesis cannot be rejected, indicating that farmers’ endowment and technology cognition do not significantly affect the adoption of green production technology among small-scale farmers, reflecting that it is more challenging to promote green production technology among small-scale farmers. This may be due to the fact that small-scale farmers are mostly subsistence farmers who engage in food cultivation mainly for subsistence purposes, and lack interest in green production technology. Under the basic condition of “big country and small farmers”, it is difficult to improve the adoption level of small-scale farmers to promote green production technology. The farmer’s endowment and technology perception variables in Model 7 are significantly positive at the 1% level, and the interaction term of farmer’s endowment and technology perception is significantly positive at the 10% level. This indicates that farmer endowment has a significant positive effect on adopting green production technologies by part-time farmers, and technology perceptions effectively play a positive moderating effect. Thus, part-time farmers should be the critical target of green production technology in agriculture. The emphasis should be on strengthening part-time farmers’ endowment and technology perception to improve their adoption of green production technology. The variables of farmers’ endowment and technology cognition in Model 8 are not significant, indicating that the effects of farmers’ endowment and technology cognition on the adoption of green production technologies by large professional households are not significant, which is not consistent with expectations. The reason may come from two aspects: on the one hand, green production technology is more prevalent among large cereal growers, and the adoption rate of green production technology among large cereal growers in the sample is less differentiated, which leads to statistically insignificant. On the other hand, large professional households have better endowments and already have higher awareness of green production technology, and their farmer endowment and technology awareness lack room for improvement.

(2) Heterogeneity analysis of farmers’ adoption behavior by technology type. Agricultural green production technologies are a collection of technologies, and the essential characteristics, complexity, and asset specificity of different technologies differ, and such differences may affect farmers’ adoption behaviors. For this reason, this paper further compares the adoption differences among farmers for different types of green production technologies. The OLS estimation was conducted with the adoption degree of green pest control technology, soil testing and fertilization technology, moderate deep loosening technology, straw returning technology and less tillage and no-tillage seeding technology as dependent variables, and the heterogeneous adoption behaviors of farmers by technology type were estimated as shown in

Table 6. Estimated results of heterogeneity of green production technologies.

	Pest and Disease Green Control	Soil Testing and Fertilization	Moderate Deep Relaxation	Straw Return to the Field	No-till Seeding with Less Tillage
Endowment	0.424 ***	0.528 **	0.898 ***	0.123	0.595 **
cognition	0.111	0.357 ***	0.066	−0.119	0.172
Endowment × cognition	0.0101	0.394 ***	−0.058	−0.172	−0.059
Transfer	−0.037	−0.521 *	−0.029	−0.051	0.471 **
Training	0.601 **	−0.575 **	0.342	0.405	0.162
Promotion	−0.361	0.351 *	−0.718 **	−0.645 **	−0.363
Subsidy	0.607 **	0.563 **	0.333	0.631 ***	0.464 *
Constant	0.294	0.458 *	−0.830 ***	0.451 **	−0.407
R2	0.082	0.120	0.126	0.028	0.105

Note: *, **, and *** indicate significant at the statistical level of 10%, 5%, and 1%, respectively.

.

From Table 6, it can be seen that farmers’ endowments have a significant positive influence on the uptake of pest and disease green control technology, soil testing and fertilizer application technology, moderate deep loosening technology, and less tillage and no-tillage technology, and has no significant influence on straw returning technology. In terms of the degree of influence, the adoption behavior of each technology was affected by farmers’ endowments in the following order: moderate deep loosening technology, less tillage and no-tillage technology, soil testing and fertilizer application technology, and green pest control technology. Moderately deep loosening technology requires specialized agricultural machinery and supporting tools to realize. Such assets have a high degree of specialization, enormous input costs, and long payback periods, At the same time, farmers with low economic capital endowments are more focused on short-term returns and tend to avoid purchasing them. The survey found that the green pest control technology is relatively mature. Farmers can purchase their green pesticides for wheat and corn, which are simple to operate, not limited by the scale of operation, and less affected by farmers’ endowments. Straw return technology is closely related to the current stage of harvesting technology upgrades in Shandong Province; wheat and corn straw harvesting and crushing recycling machines have been relatively complete in popularity, so the impact of farmer endowment on straw return technology is not significant.

The moderating effect of technology perceptions was significant for soil-formula fertilization technology but not for green pest control technology, moderately deep loosening technology, straw return technology, or less tillage and no-tillage technology. This may be due to the fact that, in the practice of grain production, soil testing and fertilizer application technology deviates from traditional agricultural production experience to a certain extent, leading to a bifurcation in farmers’ perceptions of this technology. With the expansion of a farmer’s business scale and the improvement of information access, a farmer’s awareness of soil testing and fertilizer application technology increases, and they are more inclined to adopt the technology. For this reason, we should further strengthen the popularization and promotion of green production technology and narrow the “cognitive gap” between the two technologies for large-scale farmers. The three technologies of green pest control, less tillage, and no-till technology are simple in principle and more accessible for farmers to recognize. With the gradual improvement of the grassroots promotion mechanism of agricultural green production technology, farmers in the demonstration area have relatively adequate knowledge of these three technologies, and there is little room for improvement. Even if they continue to deepen their cognitive level, it is futile to improve the adoption rate further.

4.4. Robustness and Endogeneity Tests

(1) Robustness test: In order to test the validity of the estimation results, we conducted two types of robustness tests: regression and model replacement. First, the sample was divided into two groups, with and without position, according to “whether they hold a position in the village”, and the regressions were conducted separately to compare whether there was a significant change in the results. As seen from

Table 7. Robustness and endogeneity test results.

	Sub-Sample Regression		Model Replacement	Neighborhood Emulation Tool Variables
	With Job Group	Jobless Group	Logit Regression	Resulting Equation	Select Equation
Endowment	0.317 **	0.323 ***	0.570 **	0.066 ***	0.318 ***
Transfer	0.365	0.072	−0.180	−0.016	−0.115
Training	0.531	0.572 ***	1.248 ***	0.061	0.585 ***
Promotion	−0.124	−0.139	−0.579	−0.065	−0.307
Subsidy	0.780 *	0.551 ***	1.678 ***	0.283 ***	0.910 ***
Constant	1.531 ***	1.820 ***	1.282 ***	0.533 ***	0.775 ***
R2	0.350	0.165	0.201
Wald test				38.120 ***

Note: *, **, and *** indicate significant at the statistical level of 10%, 5%, and 1%, respectively.

, the estimation results of the model with and without position group are more consistent with the estimation results of the baseline model, which verifies the robustness of the model estimation results. Further, based on the number of green production technologies adopted by farmers, the data with zero adoption were left-grouped to “0” (no adoption) and those with greater than zero adoption were right-grouped to “1” (partial or complete adoption), thus forming a binary classification dependent variable data. On this basis, the logit regression model was used as an alternative model to the Heckman model for regression in order to compare whether there was a difference in the results. As seen from Table 7, the coefficients and significance of the logit regression model analysis results are basically consistent with the baseline regression model, indicating that the original model estimation results are robust. The results obtained from the above tests all indicate that the empirical results of the benchmark regression are robust.

(2) Endogeneity test: The Heckman model can effectively address the endogeneity caused by sample selection bias. However, the adoption of green production technologies by farmers may, in turn, affect farmers’ endowments, so the possible endogeneity bias in this paper mainly comes from reverse causality. In the “carpeting” type of technology promotion activities in green production demonstration areas, farmers may have non-spontaneous adoption behaviors, increasing their economic and social capital endowments through land transfer and joining cooperatives. To this end, this study incorporates “neighborhood emulation” as an instrumental variable in the farmers’ adoption behavior model. The sign and significance of the estimated coefficients are in high agreement with the baseline regression results, indicating that farmers’ endowments can positively and significantly influence farmers’ green production technology adoption. In summary, the conclusion above is reliable.

5. Discussion and Policy Insights

5.1. Discussion

This paper aims to explore the influence of farmers’ endowments and technology perceptions on farmers’ green production technology adoption behaviors. The study found that farmers’ endowments can significantly and positively promote farmers’ adoption of green production technologies. Technology cognition can effectively promote the positive influence relationship of farmers’ endowments on green production technology adoption behaviors. This finding answers the research question.

On the endowment of the actor subject, French sociologist Bourdieu has proposed that there are four capitals or endowments, containing cultural, financial, symbolic, and social endowment [33]. It informs this paper on the definition of farmer’s endowments and their classification. Elahi et al. [10] found that endowments are positively associated with adopting new technology. Joffre et al. [11] found that farmers in social networks are more likely to adopt new technologies. This study finds the same as this paper. The psychologists Abraham and Sheeran, who studied human health problems, indicated that if a problem is considered necessary by stakeholders, then they will use the available, accessible, or not costly solution [34,35]. This paper draws on that finding. Based on the rational economic person hypothesis, this paper argues whether farmers adopt green agricultural production technologies depending on their cost–benefit ratio.

Farmers’ adoption behaviors for agricultural green production technologies should match China’s rural ecological environment management policies. Many studies have considered the impact of macro-level technology promotion policies [36,37], but few studies have incorporated the lag of policy impact into the research system of farmers’ behaviors. In the future, we will further explore the behavioral decision-making mechanisms of farmers in different policy environments to provide micro-foundational support for macro-level policy innovation. In addition, different types of farming households are different in their dependence on agriculture and their adoption behavior of agricultural green production technologies. Most existing research results regard farmers as homogeneous objects [8,15,38,39], and few studies have combined the reality of continuous differentiation of farmers for in-depth investigation. In the future, farm household differentiation should be incorporated into the behavior-driven system, abandoning the traditional “one-size-fits-all” approach [40], and providing references for the formulation of targeted agricultural policies.

5.2. Policy Insights

Given the structural imbalance in farmers’ adoption of green production technologies, this paper proposes the following policy recommendations based on the promotion of green production technology adoption by farmers’ endowments and the moderating effect of technology perception.

(1) Since the findings show that farmer endowments can promote the adoption of green technologies, we should consider the progressive enhancement of farmer endowments. The focus should be on small-scale farmers and part-time farmers with low levels of the capital endowment. First, policies should actively support the cultivation of large planters through land trust and land transfer in order to enhance the economic capital endowment of farmers. Second, they should use cooperatives and other platforms to actively carry out green production technology exchange activities in order to enhance the social capital endowment of farmers.

(2) Since the findings show that technology cognition can positively moderate the effect of farmers’ endowments on green technology adoption, we should enhance farmers’ knowledge of green production technology. First, we should seek ways and means to promote green production technology. In addition to widespread science publicity and education to enhance farmers’ technical awareness, it is necessary to construct environmentally friendly agricultural technology demonstration areas and technical guidance in the field. We should also subconsciously enhance farmers’ environmental awareness of green production technology and enhance the regulatory role of other factors. Second, in the process of green production technology promotion, technical extension personnel should change their roles and “tailor” cost–benefit analyses for farmers, highlighting the economic functions of green production technology, such as increasing yield or improving quality, so as to enhance farmers’ economic cognition of green production technology.

(3) According to the results of the heterogeneity analysis, there are differences in the adoption of green technologies by different types of farmers. Therefore, we should promote agricultural green production technology according to individual measures. We should create joint working groups with all stakeholders related to farmers. On the one hand, we should improve the social service mechanism, strengthen the technical assistance for small-scale farmers and part-time farmers, and reduce the cost of green technology adoption. On the other hand, we should consolidate the interest linkage between new agricultural business entities and small-scale farmers and enhance the ability of new entities to drive small-scale farmers.

6. Conclusions

(1) Regarding adoption numbers, more than half of the farmers adopted more than three green production technologies, and only 11.5% did not adopt any. Regarding adoption structure, the adoption rate of green pest control technology was the highest at 67.7%, while the adoption rate of moderate deep pine technology was the lowest at 32.7%. The adoption structure should be further optimized.

(2) Regarding the main effects, farmers’ endowments can significantly and positively promote farmers’ adoption of green production technologies. Among them, human capital endowment and economic capital endowment can significantly and positively influence farmers’ green production technology adoption behaviors. In contrast, social capital endowment does not significantly affect farmers’ green production technology adoption.

(3) Technology cognition can effectively promote the positive influence relationship of farmers’ endowments on green production technology adoption, but the moderating effect of technology cognition is weak, indicating that technology diffusion has a long-term nature. Compared with technology–environmental cognition, the moderating effect of technology–economic cognition is stronger, and farmers’ cognition of the economic functions of green production technology is more significant than their environmental function cognition.

(4) The results showed that farmer endowment significantly positively affected part-time farmers’ adoption of green production technologies. Technology cognition effectively played a positive regulating effect, showing that green production technology promotion should focus on part-time farmers. Most large grain farmers would consciously adopt green production technologies. Part-time farmers should be encouraged to transition to professional households or withdraw from agricultural production and operation through policy support and other means. The results of differentiating technology types showed that farmer endowment had significant positive effects on green pest control technology, soil testing and fertilizer application technology, moderate deep loosening technology, and less tillage and no-tillage technology, and it had insignificant effects on straw returning technology. The moderating effect of technology cognition had significant effects on soil testing and fertilizer application technology, and insignificant effects on other technologies.

The above findings fully explain the research aims of this paper. Including technology cognition in the analytical framework of the influence of farmers’ endowments on green production technology adoption fully considers the moderating effect of farmers’ technology cognition and enriches the existing studies on farmers’ green production technology adoption. However, farmers’ green production technology adoption behaviors are not only influenced by their endowments, but also disturbed by other factors such as the external environment, which have not been included in this paper, and the core factors of farmers’ technology adoption behaviors can be further explored from external factors in the future.