Farmers' technology preference and influencing factors for pesticide reduction: evidence from Hubei Province, China.

Based on the successful experience of pesticide reduction in China, this study uses survey data from Hubei Province to measure rice farmers' technology preferences for pesticide reduction considering their needs, and compare the heterogeneous factors influencing farmers' adoption behavior. The results show that large-scale farmers prefer drone services and efficient machinery, while small-scale farmers prefer scientific standards and biopesticides for pesticide reduction. Second, farmers' adoption behavior of pesticide reduction technologies is mostly influenced by education, risk attitude, income, agricultural labor, scale, rice price, residue testing, brand, training, subsidy, and demonstration. Among them, education, risk attitude, scale, rice price, cost, and training, significantly affect farmers' adoption level of multiple pesticide reduction technologies. Further, higher rice prices and participation in training could promote the use of pesticide reduction technologies in a larger area by farmers. Therefore, the real needs of farmers should be focused on the promotion of pesticide reduction technologies, and pesticide reduction programs in different regions should carry out precise intervention policies. These findings can provide practical policy guidance for effective pesticide reduction in the central region of China.

Introduction

Excessive and inefficient use of chemical pesticides is common in many developing countries, although the use of pesticides is still an important measure of pest management (Sun et al. 2020; Huang et al. 2021). For a long time, humans have used chemical pesticides in large quantities to stabilize and improve crop yields and solve basic survival problems. But, the characteristics of chemical pesticides which are high toxicity, easy residue, and difficult degradation, not only seriously endanger human health, but also cause ecological damage and environmental pollution (Gould et al. 2018; Luo et al. 2021). Therefore, how work on chemical pesticide reduction is the common goal of countries around the world to achieve sustainable, low-carbon, and green development in agriculture in recent decades (Wossink 2007; Lévesque et al. 2021). For example, Japan has developed a hierarchical Japanese Agriculture Standard, which has greatly encouraged farmers to use as few pesticides as possible. Australia is early to realize the development of organic farming and carbon sink farming, and agricultural subsidies are linked to agricultural environmental assessment in Europe.

To achieve pesticide reduction goals, scholars have made valuable arguments about how to incentivize agricultural producers to use fewer pesticides. On the one hand, scholars have focused on increasing the adoption of biopesticides, green control technologies, and integrated pest management (IPM) by farmers to replace the use of chemical pesticides. For example, Petrescu-Mag et al. (2019) investigated farmers’ willingness to pay for biopesticides. Gao et al. (2017) explored the adoption of green control technologies in Chinese family farms, and Wyckhuys et al. (2021) also call for the worldwide insistence on promoting IPM in agriculture to achieve pesticide reduction. On the other hand, scholars have explored the factors that lead to the overuse and misuse of pesticides by farmers. Yang et al. (2019) found that unfriendly green agricultural markets and imperfect policies are the main driving factors of pesticide overuse for farmers. Sun et al. (2019) found that the lack of individual knowledge and ability of farmers and the imperfect agricultural extension services are important reasons for farmers’ misuse of pesticides. The above research undoubtedly plays an important role in guiding the reduction of pesticides in agricultural production practices around the world.

China, the world’s largest producer and consumer of pesticides, at the same time also made greater achievements in pesticide reduction after 2015. The data1 show that the growth rates of total pesticide use in China in 2015, 2016, 2017, 2018, and 2019 were − 1.13%, − 2.41%, − 4.89%, − 9.12%, and − 6.72%, actually achieving “negative growth.” The effect of pesticide reduction in China is very significant. Since 2012, the Chinese government has attached great importance to ecological and environmental issues and formulated a greening development strategy, and in 2015, the Ministry of Agriculture and Rural Affairs implemented the “Zero Growth Action Plan for Pesticide Use by 2020” to reduce the use of pesticides, and some forecast data show that the rate of pesticide reduction in China will continue to be maintained in the coming decades. Therefore, China’s successful experience in controlling the growth of pesticide use rapidly and effectively has some practical reference value.

With the success of pesticide reduction in China since 2015, this study uses data from a survey of rice farmers in Hubei Province, China to focus on answering two questions: which pesticides reduction technologies do farmers prefer? What factors influence farmers’ pesticide reduction technology adoption decisions? Distinguishing from the previous literature, the main contributions of this paper are as follows: First, many previous studies have focused on the supply side of agricultural extension and have only isolated the adoption behavior of farmers for a single pesticide reduction technology, which ignores the real needs of farmers themselves. Because there are many pesticide reduction technologies to choose from in practice, and farmers’ technology adoption preferences are heterogeneous (Bazoche et al. 2011; Lévesque et al. 2021). For example, it is not mandatory for all farmers to adopt biopesticides. Therefore, The Weighted Frequency Approach has been used to measure the priority order of farmers’ technology adoption preferences for pesticide reduction. Second, different pesticide reduction technologies have been applied to different groups of farmers, and thus there are potential differences in the factors influencing their adoption. We then use the Heckman model to compare the factors influencing the adoption of various pesticide reduction technologies to improve the applicability of the technologies. The answers to the above two questions are beneficial to the development of actionable regional pesticide reduction plans based on respect for the true wishes of farmers.

Materials and methods

Survey data

Affected by the COVID-19, cross-provincial rural surveys are regulated in China. Therefore, we chose Hubei Province in China as our sampling area, which is both the province where our institution is located, and representative rice-producing region in central China. Since 2015, the agricultural affairs department of Hubei Province is issuing guidance programs to farmers on the implementation of pesticide reduction technologies for major food crops. The pesticide reduction technology has also been made a key component of agricultural extension. So, Hubei Province is a representative region for our study of pesticide reduction in China. We selected a total of ten cities (counties) from the major rice-growing regions in Hubei Province in July 2021, according to the principle of stratified random sampling in Xiantao, Qianjiang, Jingshan, Shayang, Zaoyang, Xiangzhou, Jianli, Shishou, Xishui, and Tuanfeng (Fig. 1). Then, 10–20 villages are randomly selected from each city (county), and 10–15 rice growers are selected from each village for investigation. We finally succeeded in obtaining 1193 valid data that could be used for empirical analysis. It is worth noting that in order to make our samples more representative and practical reference, the selection of study sample areas refers to the official list of green control pilot counties2 published by the Chinese Ministry of Agriculture and Rural Affairs. The survey questionnaire mainly included farmers’ personal characteristics, family characteristics, and production management characteristics. Among them, the focus was on collecting detailed data on rice farmers’ pesticides inputs and reduction technologies adoption.

Fig. 1

Distribution of study areas. Notes: Different colors represent the geographic locations where samples are obtained, and only up to the city (county) level are indicated here

Model

The Weighted Frequency Approach (WFA) has been used to measure the priority order of rice farmers’ technology adoption preferences for pesticide reduction. First, we need to define the categories of pesticide reduction technologies. Drawing on the official document “Zero Growth Action Plan for Pesticide Use by 2020” issued by China’s Ministry of Agriculture and Rural Affairs. Pesticide reduction techniques are grouped into four categories: Control, Replacement, Precision, and Unification (CRPU). See Table 1 for a detailed description.

Table 1

Categories of pesticide reduction technologies in China

Category	Definition	Technology
Control	That is, to achieve sustainable control of pesticides. Promote green control techniques to create environmental conditions conducive to crop growth, natural enemy protection, and pests suppression	• Ecological control • Physical trapping
Replacement	That is, biopesticides instead of chemical pesticides, efficient machinery instead of small inefficient machinery. The purpose is to expand the use of low-toxicity pesticides and improve the efficiency of pesticide use	• Biopesticides • Efficient machinery
Precision	That is, to achieve the precise use of pesticides. The focus is on the accurate identification of pests and the use of suitable doses of pesticides at the right time. Avoid farmers misusing pesticides	• Scientific standard
Unification	That is, to achieve unified pest control. Support specialized service organizations for pest control, to solve the difficulties of scattered smallholder farmers’ pesticide use	• Drone service

Official policy texts are available at http://www.zzys.moa.gov.cn/gzdt/201503/t20150318_6309945.htm. The “Technology” column in the table only lists some typical examples of pesticide reduction technologies in rice cultivation, but not all. Biological control specifically refers to the ecosystem consisting of rice, ducks, lobsters, and frogs. Efficient machinery refers to large electric or oil pesticide spraying machinery. The substitution of biopesticides for chemical pesticides is also considered to be an effective means to reduce the total amount of pesticides used. The pesticide reduction emphasized in this paper includes biopesticides and chemical pesticides

Second, pesticide reduction technology preference ranking was collected from rice farmers. The questionnaire was designed with items to allow each farmer to rank the above six technologies in order of their need from 1–6 (1 = strong preference and 6 = no preference at all). Finally, the cumulative sample size of each pesticide reduction technology in order of 1–6 is used to measure the composite score of demand preference for each technology. The calculation formula is as follows:where, represents the composite score of demand preference for the th technology. and are the cumulative frequencies and weights for the th ranking of this technology, respectively. To avoid the errors associated with individual subjective assignments, we use equidistant assignments. That is, . Then, the ascending order of is used to determine the priority of farmers’ pesticide reduction technology preferences.

The Heckman model was used to estimate the factors affecting the adoption of pesticide reduction technologies among rice farmers. This model can not only argue the factors influencing farmers’ adoption (or not) of pesticide reduction technology, but also analyze the factors influencing the level (proportion) of technology adoption, and the Heckman model can deal with the potential self-selection and endogeneity problems (Heckman and Richard 1985). First of all, we construct a model of the factors influencing pesticide reduction technology adoption among rice farmers based on previous scholarly research (Fig. 2).

Fig. 2

Model of factors influencing farmer’s adoption of pesticide reduction technologies. Notes: The figure refers only to the listing of the main influencing variables

These factors mainly contain personal factors, family factors, production management factors, market factors, and policy factors (Grovermann et al. 2017; Pan et al. 2021; Zhao et al. 2021). As for personal factors, we selected four important indicators: age, gender, education, and risk attitude. These factors are considered to influence the production habits and cognitive level of farmers, which further affects their pesticide use decisions (Khan and Damalas, 2015; Huang et al. 2021). As for family factors, income, agricultural labor, and children were chosen. Among them, income and children members of the family influenced farmers’ consumer demand for high-quality agricultural products (Benoît et al. 2020). Agricultural labor is also the key to influencing the efficiency of pesticide use (Guo et al. 2020), and the scale and cooperative organization were chosen as production management factor variables because of the presence of scale effects (Qin and Lv 2020). In addition, market factors are directly related to the production benefits of farmers and are the source of farmers’ decisions. For example, residue testing could reveal whether an agricultural product exceeds pesticide standards, and branding can effectively convey information about product quality (Yang et al. 2019). Similarly, agricultural extension policies have a non-negligible role in guiding and regulating the adoption of pesticide reduction techniques by China farmers (Sun et al. 2020).

Then, we need to construct a two-stage Heckman model using the above variables. The first stage was to use a Logit model to estimate the factors that whether rice farmers adopt pesticide reduction technologies.where, and p denote the behavior and probability of adoption of pesticide reduction technologies by farmers th, respectively. is the impact coefficient of each factor to be estimated. is the intercept term and is the random error of the model. In further estimating the factors influencing the degree of adoption of pesticide reduction technologies, the problem of self-selection bias of the sample needs to be addressed. We need to obtain the inverse Mills ratio in the first stage estimation, obtained by dividing the density function of the normal distribution with the cumulative distribution function.

Then, the second stage equation has been constructed substituting as the independent variable and using OLS to estimate the factors influencing the level of farmers’ adoption of pesticide reduction technologies.where, is the level of farmers’ adoption of pesticide reduction technologies. That is, , and refers to the number of pesticide reduction technologies adopted by the th rice farmer. Similarly, is the impact coefficient of each factor to be estimated. is the intercept term and is the random error. The definitions and statistical descriptions of all variables in the model are shown in Table 2.

Table 2

Definition and description of variables

Variables	Definition and assignment	Average	S.D
Dependent variables
Adoption behavior	Whether the rice farmer adopts any pesticide reduction technology in CRPU: Yes = 1, no = 0	0.816	0.102
Adoption level	The level of farmer’s adoption of pesticide reduction technologies	0.122	0.059
Independent variables
Age	Age of the interviewee (year)	58.695	10.320
Gender	Gender of the interviewee: Male = 1, female = 0	0.718	0.107
Education	Number of years of education of the interviewee (year)	7.325	3.008
Risk attitude	Risk preference of interviewees: Risk averse = 1, neutral = 2, risk like = 3	1.625	0.354
Income	Total income of family members in 2020 (thousand yuan)	100.452	8.733
Agricultural labor	Number of laborers engaged in agricultural production in the household	1.748	0.592
Children	Are there any children under 6 years old in the household: Yes = 1, no = 0	0.312	0.127
Scale	The scale of farmer’s rice cultivation (ha)	0.288	0.096
Organization	Whether to participate in professional farmer cooperative organizations of rice: Yes = 1, no = 0	0.219	0.094
ice price	The average price of rice marketed for sale in 2020 (yuan/kg)	2.171	0.282
Cost	The average cost of rice pest control in 2020 (thousand yuan/ha)	1.282	0.234
Residue testing	Whether the rice sold is tested for pesticide residues: Yes = 1, no = 0	0.074	0.009
Brand	Whether the produced rice has a brand: Yes = 1, no = 0	0.061	0.028
Training	Whether to participate in technical training on pesticide reduction: Yes = 1, no = 0	0.433	0.150
Subsidy	Whether to receive subsidies for adopting pesticide reduction technologies: Yes = 1, no = 0	0.150	0.082
Demonstration	Is there a demonstration base of pesticide reduction technology in the vicinity: Yes = 1, no = 0	0.378	0.109
Region	Sample farmers belong to the study area: Jingzhou = 1, others = 0	0.101	0.435

The data in the table are counted from 1193 survey questionnaires. 1 ha = 15 mu in China. Due to the differences in economic and cultural levels in different regions, we control the regional variables in the form of virtual variables. Here, we only take Jingzhou as an example. The indicator of subsidy refers to government subsidies for farmers to adopt pesticide reduction technologies in order to achieve pesticide reduction policy goals. For example, the government of Shishou City subsidizes the physical trapping technology adopted by farmers, and Qianjiang City subsidizes the purchase of green biopesticides for some farmers

Results and discussion

Results of technology preference prioritization

We obtained the pesticide reduction technology adoption preference priority order of rice farmers using the WFA. The results can be seen in Fig. 3. First, the sample statistics from the 1–6th order of the single technology demand preference in Fig. 3(a) show that the scientific standard is placed in an important position, with the largest sample size in the first and second positions. In the actual survey, farmers have elaborated their willingness to follow pesticide labels for the scientific use of pesticides in order to meet the national policy of pesticide reduction. Second, the composite scores of the pesticide reduction technologies in Fig. 3(b) are in descending order: Scientific standard > Biopesticides > Efficient machinery > Ecological control > Physical trapping > Drone services. In other words, rice farmers’ preferences for the four categories of pesticide reduction techniques were in the order of precision, replacement, control, and unification. That is, following scientific application standards is still the most preferred method of pesticide reduction by rice farmers. Third, we grouped statistics on pesticide reduction technology preferences of rice farmers of different scales and found differences. Large-scale farmers prefer to select drone service and efficient machinery to achieve pesticide reduction. While small-scale farmers prefer to select scientific standards and biopesticides.

Fig. 3

Pesticide reduction technology preference scores of samples. Notes: Figure (a) is the distribution of preference scores for pesticide reduction technologies. (b) is the composite score of each technology calculated by WFA. Different weights will be given to each technology in the process of score calculation. And the full sample is divided into two subsamples, large-scale and small-scale, according to the 1 ha threshold for recounting

The above results are valuable for guiding the reduction of pesticides in practice. On the one hand, it is fundamental to follow scientific standards for pesticide application, especially for small-scale farmers. Many scholars have pointed out that overuse, misuse, and ineffective use of pesticides due to non-compliance with scientific standards of pesticide labels on the timing, dosage, variety, and function of pesticides are common in developing countries (Sun et al. 2019; Huang et al. 2021; Bagheri et al. 2021). Due to the limitations of farmers’ own literacy, cognitive level, and pesticide knowledge skills, farmers do not fully rely on their own production experience to effectively control pests (Petrescu-Mag et al. 2019). Farmers’ own experience is even called a “bad habit” for pesticides application (Wang et al. 2018). On the other hand, differentiated pesticide reduction technology extension activities should be carried out to meet the real needs of farmers of different scales. Large-scale farmers have a large demand for drone services and efficient machinery out of the need to save labor costs and improve production efficiency (Qin and Lv 2020). In contrast, small-scale farmers need to bear larger transaction and negotiation costs in using large machinery (Gao et al. 2021).

Estimation results of factors influencing the adoption of pesticide reduction technology

The variables defined above were put into Eq. (2) in order to explore the factors affecting farmers’ choice of pesticide reduction technologies. First, we tested the variables in the model for covariance by the variance inflation factor values to rule out possible serious covariance problems among the variables. Next, the coefficients of the model were estimated in Stata software using Logit robust estimation to avoid heteroskedasticity problems due to model omission of important variables or sample outliers. The factors influencing the adoption behavior of the six pesticide reduction technologies among the sample farmers were regressed in turn (Table 3), and the estimated results of the model coefficients were more reliable as indicated by the overall test Wald values of the regression results of each model.

Table 3

Estimation of factors influencing farmers’ adoption of pesticide reduction technology

Variables	M1: Ecological control	M2: Physical trapping	M3: Biopesticides	M4: Efficient machinery	M5: Scientific standard	M6: Drone service
	Coefficient (S.E.)	Coefficient (S.E.)	Coefficient (S.E.)	Coefficient (S.E.)	Coefficient (S.E.)	Coefficient (S.E.)
Age	− 0.290 (0.275)	0.023 (0.165)	− 0.454 (0.266)*	0.725 (0.518)	− 0.314 (0.119)**	0.581 (0.413)
Gender	− 0.015 (0.019)	0.052 (0.040)	− 0.294 (0.584)	− 0.613 (0.509)	− 0.135 (0.043)**	0.107 (0.036)
Education	0.140 (0.050)**	0.071 (0.048)	0.030 (0.015)**	− 0.002 (0.012)	0.312 (0.083)***	0.131 (0.074)*
Risk attitude	0.037 (0.012)***	0.197 (0.026)***	− 0.047 (0.043)	− 0.106 (0.133)	− 0.168 (0.086)*	− 0.290 (0.359)
Income	0.029 (0.016)*	0.017 (0.014)	0.358 (0.173)**	0.121 (0.056)**	0.048 (0.037)	0.045 (0.022)**
Agricultural labor	0.591 (0.415)	0.554 (0.157)***	0.037 (0.028)	− 0.030 (0.014)**	0.034 (0.673)	− 0.849 (0.325)**
Children	− 0.002 (0.002)	0.002 (0.003)	0.088 (0.008)***	− 0.002 (0.005)	0.019 (0.009)**	0.004 (0.006)
Scale	0.339 (0.248)	− 0.302 (0.173)*	− 0.121 (0.211)	0.169 (0.046)***	0.166 (0.465)	0.333 (0.028)***
Organization	0.540 (0.524)	0.659 (0.513)	1.581 (0.622)	2.040 (0.481)***	0.360 (0.277)	0.374 (0.169)
Rice price	0.242 (0.051)***	− 0.163 (0.338)	0.277 (0.258)	− 0.156 (0.271)	1.015 (0.509)*	− 0.970 (1.092)
Cost	− 0.141 (0.076)*	− 0.730 (1.135)	0.005 (0.016)	− 0.127 (0.071)*	− 0.036 (0.563)	0.043 (0.526)
Residue testing	3.360 (0.376)	− 5.467 (7.865)	2.685 (0.877)***	4.211 (4.441)	4.885 (2.414)**	1.480 (1.504)
Brand	0.195 (0.109)*	− 0.196 (0.136)	0.130 (0.054)**	− 0.108 (0.093)	0.228 (0.078)**	4.157 (7.006)
Training	0.115 (0.049)**	0.175 (0.027)***	0.389 (0.115)***	0.262 (0.354)	0.148 (0.035)***	− 0.029 (0.257)
Subsidy	0.396 (0.247)	0.411 (0.150)**	0.420 (0.181)	0.309 (0.144)**	− 0.047 (0.378)	0.071 (0.016)***
Demonstration	0.283 (0.156)*	0.629 (0.357)*	0.408 (0.227)**	0.153 (0.226)	0.558 (0.795)	0.956 (0.390)**
Region	0.275 (0.136)**	− 1.964 (2.602)	0.691 (0.595)	− 3.211 (1.471)**	− 0.643 (1.098)	− 0.408 (0.190)*
Pseudo R2	0.085	0.082	0.057	0.165	0.176	0.129
Wald Chi2	28.062***	27.093***	60.562**	26.370***	23.054***	17.985**

Standard errors of coefficients are in parentheses. The constant term results are not presented. *, **, and *** indicate significance at the statistical levels of 10%, 5%, and 1%, respectively

Farmer characteristics variables affecting adoption of pesticide reduction technologies. There is a significant negative effect of interviewee age on the adoption of biopesticides and scientific standards. This is because the older the farmer is, the more profoundly influenced by the traditional crude chemical pesticide use habits (Bagheri et al. 2021). The effect of gender on scientific standards is significantly negative. It indicates that women are more concerned about the normative use of pesticides attributed to their concerns about the quality of food (Khan and Damalas 2015). The effect of education on the adoption of multiple pesticide reduction technologies is significantly positive. This validates the emphasis of Yang et al. (2014) on the importance of education in guiding the regulation of pesticide use by farmers. Risk attitude had a positive effect on ecological control and on physical trapping, but a negative effect on scientific standard. Ecological and physical control are typical emerging green control technologies, which require “adventurers” to make brave attempts (Gao et al. 2017). In contrast, risk-averse farmers are more likely to use chemical pesticides beyond standards (Salazar and Rand 2020). Similarly, farmers in high-income households are more likely to reduce pesticides. The results also show that families with children are more likely to use biopesticides and follow scientific standards. Because they have a greater demand for high-quality agricultural products. This is to protect their own and their families health from pesticide poisoning (Benoît et al. 2020). In addition, the scale has a significant positive impact on the adoption of efficient machinery and drone service. This suggests that such socialized services are more likely to be favored by large-scale farmers, and large and efficient pesticide spraying machinery is mostly occupied by cooperative organizations in practice. This shows that pesticide reduction is more likely to be achieved by promoting efficient machinery and drone services to cooperatives and large-scale farmers (Huang et al. 2021).

Market variables affecting farmers’ adoption of pesticide reduction technologies. Rice prices have a significant positive impact on rice farmers’ adoption of ecological control and scientific standards. Rice produced with fewer pesticides is given the attribute characteristic of “green,” which could achieve higher market prices and consumer favors (Yang et al. 2019). Of course, some high-value rice production also sets strict production standards for the use of chemical pesticides (Sun et al. 2020). Pest management costs have a significant negative impact on rice farmers’ adoption of ecological control and efficient machinery. This is because ecological control and the introduction of efficient machinery require farmers to bear high costs, such as the purchase of specialized equipment, raw materials, and fuel (Gao et al. 2021). Residue testing has a significant positive impact on rice farmers’ adoption of biopesticides and scientific standards. Pesticide residue testing is the main means of identifying the quality and safety of agricultural products. Importantly, the use of highly toxic chemical pesticides is the main cause of pesticide residues (Gould et al. 2018), and the use of pesticides in accordance with scientific standards, or the use of biopesticides can undoubtedly reduce pesticide residues. In addition, agricultural product brands benefit rice farmers in adopting ecological control, biopesticides, and scientific standards. The construction of agricultural product brands needs to follow certain production standards, especially the prohibition of the use of highly toxic chemical pesticides. This can not only increase the premium of the agricultural product brand, but also maintain its long-term reputation of the brand (Li and Guo 2019).

Policy variables affecting farmer’s adoption of pesticide reduction technologies. Technical training has a significant positive impact on the adoption of pesticide reduction technologies for rice farmers, in addition to efficient machinery and drone services. By conducting technical training and guidance, agricultural extension personnel can effectively solve the difficulties encountered by farmers in the actual application of pesticide reduction technology and maximize the guarantee of technical effects (Wuepper et al. 2021). In addition, technical training is also conducive to the improvement of farmers’ pesticide awareness levels and the accumulation of pest and disease knowledge and skills (Sun et al. 2019). Subsidies have a significant positive impact on the rice farmers’ adoption of physical trapping, efficient machinery, and drone services. This is due to the Chinese government’s strong support and promotion of efficient machinery and drone services in recent years. Policy subsidies can effectively reduce the acquisition cost of pesticide reduction technologies and weaken the probability of risk loss from the uncertainty of technology use on economic returns (Li et al. 2018). The technology demonstration has a significant positive impact on the rice farmers’ adoption of biological control, physical trapping, biopesticides, and drone services. Because technology demonstration for farmers can directly observe the use of the technology and its economic benefits, provide reference advice to neighboring farmers, and eliminate the “worries” in farmers’ minds (Pan et al. 2021).

Finally, it should be emphasized that regional variables also have a significant impact on the rice farmers’ adoption of ecological control, efficient machinery, and drone services. This shows that many of China’s pesticide reduction technology promotion methods have typical regional characteristics. That is to say, the focus of technology promotion in different regions is different. Combined with the reality of our survey, most hills similar to Huanggang City in Hubei Province, China, are not suitable for the promotion of large machinery and drone service technology. In contrast, areas with many rivers compared to plains are suitable for promoting ecological control, such as the rice and lobster co-farming model in Qianjiang City, Hubei Province.

Estimation results of the factors influencing farmer’s adoption level of pesticide reduction technologies

To explore how the above variables further influence the level of pesticide reduction technology adoption among rice farmers, Heckman regressions on the sample were used (Table 4). Since the presence or absence of subsidies only addresses whether the adoption behavior occurs, it was removed in the second stage of the regression, thus satisfying the basic requirements of the Heckman regression. The significance of the inverse Mills ratio obtained from the regression indicates that the problem of sample selectivity bias exists and that the Heckman two-stage regression method is reasonably applied.

Table 4

Estimation results of the Heckman model and its robustness test

Variables	First stage:Adoption behavior	Second stage:Adoption level	Tobit 1:Adoption level	Tobit 2:Area ratio
	Coefficient (S.E.)	Coefficient (S.E.)	Coefficient (S.E.)	Coefficient (S.E.)
Age	0.004 (0.003)	0.153 (0.172)	0.007 (0.005)	0.002 (0.003)
Gender	0.142 (0.151)	0.114 (0.061)	0.002 (0.005)	0.003 (0.003)
Education	0.228 (0.094)**	0.045 (0.022)**	0.630 (0.247)**	0.098 (0.042)**
Risk attitude	0.183 (0.106)*	0.609 (0.326)*	0.051 (0.028)**	0.007 (0.002)***
Income	0.096 (0.039)**	0.003 (0.022)	0.002 (0.001)*	0.001 (0.001)
Agricultural labor	0.019 (0.010)*	1.056 (0.813)	0.510 (0.613)	0.086 (0.051)*
Children	0.013 (0.015)	0.004 (0.006)	− 0.004 (0.003)	− 0.004 (0.003)
Scale	0.237 (0.126)*	0.248 (0.055)***	0.227 (0.096)**	− 0.067 (0.040)*
Organization	0.757 (0.719)	0.215 (0.742)	0.004 (0.003)	0.001 (0.001)
Rice price	0.291 (0.122)**	0.144 (0.081)*	0.011 (0.007)*	0.022 (0.009)**
Cost	− 0.194 (0.626)	− 0.136 (0.062)**	− 0.004 (0.002)*	− 0.003 (0.002)*
Residue testing	3.692 (2.111)*	4.153 (3.092)	0.002 (0.002)	− 0.002 (0.002)
Brand	1.349 (0.635)**	− 1.993 (1.147)	0.182 (0.111)*	0.018 (0.132)
Training	0.187 (0.055)***	0.205 (0.088)**	0.041 (0.011)***	0.097 (0.041)**
Subsidy	0.121 (0.051)**	_	− 0.004 (0.002)**	0.001 (0.001)
Demonstration	0.606 (0.343)*	0.617 (0.522)	0.016 (0.182)	0.012 (0.015)
Region	− 0.396 (1.085)*	− 0.487 (0.872)	− 0.009 (0.004)**	0.005 (0.004)
φ	_	0.802 (0.371)**	_	_
Pseudo R2	0.194	0.168	0.021	0.018
Wald Chi2	20.784***	18.740***	_	_
LR Chi2	_	_	32.25***	65.72***

The estimated adoption behavior in the first stage is the adoption of any of the CRPU technologies by rice farmers. Standard errors of coefficients are in parentheses. The constant term results are not presented. *, **, and *** indicate significance at the statistical levels of 10%, 5%, and 1%, respectively. The Tobit model on the right-hand side is mainly used to test the robustness of the empirical results

First, we can obtain the factors influencing the adoption behavior of pesticide reduction technologies by rice farmers estimated from the first stage. These mainly include education, risk attitude, income, agricultural labor, scale, rice price, residue testing, brand, training, subsidy, and demonstration. Statistics on the significance of the coefficients show that they have influenced the adoption of pesticide reduction techniques by rice farmers, and these results are consistent with Table 3. So, we will not repeat the reason here.

Second, we can obtain the factors influencing the adoption level of pesticide reduction technologies by rice farmers from the second stage. The positive impact of education on the level of adoption is tested by significance. This shows that more education can promote farmers to actively adopt more pesticide reduction technologies. Risk attitudes can also significantly promote the adoption of more pesticide reduction technologies, as the adoption of many emerging technologies is considered a risky thing by farmers (Bagheri et al. 2021). The large-scale the farmer, the more types of pesticide reduction techniques they use. The economic effects of different reduction techniques need to be verified one by one in practice, so that farmers may be willing to accept a certain technology on a large scale. Otherwise, they will face a greater risk of loss (Han et al. 2021). Rising rice prices are helping farmers to adopt more pesticide reduction techniques. This is the fundamental goal of production that should be carried out by farmers in order to obtain market profits. Of course, the results also show that the cost of pest management has increased significantly after the adoption of more pesticide reduction techniques. This is inevitable because of the improvement of many raw materials and production processes (Gao et al. 2021). Of course, technical training is still necessary, if you want to make different pesticide reduction technologies quickly promoted. Different technical operation points need guidance and help from an agricultural technology extension personnel (Huang et al. 2021). This finding could provide recommendations to guide the diffusion of diversified pesticide reduction techniques.

Robustness estimation

In order to verify the stability of the above empirical results, we combined the two-step decision-making problems of pesticide reduction technology adoption behavior and the adoption level of farmers into one step. It is also necessary to set the adoption level of 0 for samples that have not adopted pesticide reduction techniques. The Tobit model, also known as the restricted dependent variable model, is suitable for handling sample selection and data consolidation problems and can be used as a robustness test in this paper. We used two approaches to obtain robust estimates, the results of which are shown in Table 4.

On the one hand, we only replace the method and estimate the influencing factors of farmers’ adoption level of pesticide reduction technologies, and find that the significant influencing factors in the regression results are basically consistent with the results obtained by Heckman. This proves the reliability and stability of the empirical results. On the other hand, we replace the dependent variable for re-estimation based on the Tobit model. That is the ratio of the area applied by pesticide reduction technology to the total area cultivated. The results show that education, risk attitudes, agricultural labor, scale, rice prices, cost, and training are important factors affecting the ratio of area adopted by farmers in pesticide reduction techniques. Among them, education and risk attitudes are important factors at the individual level. The promotion of pesticide reduction work requires highly educated and risk-taking people as pioneers (Salazar and Rand 2020). Moreover, rice prices and training are important external environmental factors. It is also necessary to drive rice farmers to widely apply pesticide reduction technology through the dual power of the market and the government (Yang et al. 2019; Pan et al. 2021). In addition, we obtained the results shown in Table 5 using the four technology categories in Table 1 as dependent variables in turn for Logit estimation. The significance of each variable in these estimation results is basically consistent with Table 3, which also verifies the robustness of the empirical results of this study, and the analysis of results will not be repeated.

Table 5

Estimation of factors influencing pesticide reduction technology category selection behavior of rice farmers

Variables	M7: Control	M8: Replacement	M9: Precision	M10: Unification
	Coefficient (S.E.)	Coefficient (S.E.)	Coefficient (S.E.)	Coefficient (S.E.)
Age	0.075 (0.028)	0.272 (0.153)*	− 0.314 (0.119)**	0.581 (0.413)
Gender	0.021 (0.030)	− 0.440 (0.585)	− 0.135 (0.043)**	0.107 (0.036)
Education	0.102 (0.060)*	0.019 (0.010)*	0.312 (0.083)***	0.131 (0.074)*
Risk attitude	0.127 (0.019)***	− 0.069 (0.107)	− 0.168 (0.086)*	− 0.290 (0.359)
Income	0.019 (0.011)*	0.327 (0.146)**	0.048 (0.037)	0.045 (0.022)**
Agricultural labor	0.576 (0.274)**	− 0.011 (0.006)*	0.034 (0.673)	− 0.849 (0.325)**
Children	0.001 (0.002)	0.074 (0.010)***	0.019 (0.009)**	0.004 (0.006)
Scale	0.002 (0.156)	0.068 (0.034)**	0.166 (0.465)	0.333 (0.028)***
Organization	0.607 (0.520)	1.873 (0.715)**	0.360 (0.277)	0.374 (0.169)
Rice price	0.103 (0.049)**	0.021 (0.354)	1.015 (0.509)*	− 0.970 (1.092)
Cost	− 0.251 (0.147)*	− 0.002 (0.001)*	− 0.036 (0.563)	0.043 (0.526)
Residue testing	− 1.821 (3.591)	2.756 (1.334)**	4.885 (2.414)**	1.480 (1.504)
Brand	0.001 (0.098)	0.031 (0.019)*	0.228 (0.078)**	4.157 (7.006)
Training	0.146 (0.061)**	0.216 (0.127)*	0.148 (0.035)***	− 0.029 (0.257)
Subsidy	0.407 (0.193)**	0.118 (0.052)**	− 0.047 (0.378)	0.071 (0.016)***
Demonstration	0.425 (0.236)*	0.254 (0.156)*	0.558 (0.795)	0.956 (0.390)**
Region	0.107 (0.063)*	− 1.119 (0.856)	− 0.643 (1.098)	− 0.408 (0.190)*
Pseudo R2	0.089	0.358	0.176	0.129
Wald Chi2	25.419***	50.260***	23.054***	17.985**

*, **, and *** indicate significance at the statistical levels of 10%, 5%, and 1%, respectively. Since the technical practices in the two technical categories of precision and unification are scientific standard and drone service, respectively, the results for M9 and M10 are the same as those for M5 and M6 in Table 3

Conclusion and policy implications

Based on the successful experience of pesticide reduction among rice farmers in Hubei Province, China, we used survey data to measure farmers’ preferences for pesticide reduction technologies and their influencing factors. This study provides a more comprehensive focus on the real needs of farmers, and these findings can provide practical policy guidance for effective pesticide reduction in developing countries. The main findings of the paper are as follows:

First, the general pesticides reduction technology adoption preferences of rice farmers are in descending order of scientific standard, biopesticides, efficient machinery, ecological control, physical trapping, and drone service. However, they also exhibit scale heterogeneity, showing that large-scale farmers prefer drone services and efficient machinery, while small-scale farmers prefer scientific standards and biopesticides. Second, the adoption behavior of different pesticide reduction technologies has different influencing factors that are mostly influenced by variables such as education, risk attitude, income, agricultural labor, scale, rice price, residue testing, brand, training, subsidy, and demonstration. Among them, education, risk attitude, scale, rice price, cost, and training are the variables that significantly affect farmers’ adoption level of multiple pesticide reduction technologies. Further, personal factors similar to education and risk attitude, and external factors similar to rice price and training combine to affect the ratio of area adopted by farmers for pesticide reduction technologies.

The following policy insights can be derived from the above findings: First, it is recommended to develop a pesticide reduction technology extension program that can meet the real needs of farmers. We should pay attention to information about the needs of farmers with different characteristics themselves and carry out personalized and precise technical extension services, specifically, for large-scale farmers to promote efficient machinery and drone services, and for small-scale farmers to promote scientific application techniques and biopesticides. Second, market and policy factors also need to be optimized, in addition to farmer-specific factors, to achieve widespread adoption of pesticide reduction technologies. It is recommended to build a green agricultural supply system to achieve high prices for quality agricultural products, and to induce farmers to take the initiative to reduce pesticide application. At the same time, strengthen pesticide reduction technology training, subsidies, demonstration, and other promotional activities to make the technology more easily understood and accepted by farmers. Therefore, it is still necessary to carry out targeted technical extension interventions in the context of designing pesticide reduction technology programs with local characteristics.