Biopesticide Extension and Rice Farmers' Adoption Behavior: A Survey From Rural China


 Although the beneficial effects of the agricultural extension of farmers’ biopesticide adoption have been largely demonstrated, the questions of what approaches can better extend biopesticides and how to improve the inefficiencies of biopesticide extension still need to be explored. In a survey of 1148 rice farmers in Hubei Province, China, the technology supply and demand theory was used to explain the low efficiency of biopesticide extension. The endogenous switching probit model was used to estimate the impact of biopesticide technology publicity, training, demonstration and subsidies on farmers’ adoption. The results show that biopesticide extension can promote rice farmers’ adoption probability by 10.3% ~ 11.7%. Among these methods, technology demonstration is currently the best way to extend biopesticides. Moreover, inadequate supply and demand of biopesticides are important for explaining the inefficiency of biopesticide extension in China. Extending biopesticides is better for farmers with smaller scales, younger ages, and lower education and for those who are cooperative members. Therefore, we should not only actively conduct biopesticide demonstration but also more importantly induce farmers’ biopesticide demand and secure the market supply of biopesticide products. These findings will provide useful guidance for biopesticide extension and pesticide reduction in China and other developing countries.


29
Biopesticides are considered necessary elements to replace chemical pesticides and realize the 30 sustainable development of agriculture (Constantine et al., 2020). They have excellent technical 31 attributes, such as low toxicity, low residue, and environmental friendliness (Srinivasan et  users do not follow implementation standards, it is difficult to achieve the desired effect of pest 47 management (Bagheri et al., 2021). Agricultural extension can deliver biopesticide technology 48 information to farmers, compensating for the misuse of pesticides caused by information 49 asymmetry (Yang et al., 2014;Huang et al., 2021). Some previous studies have also shown that 50 usefulness of the technology, and effectively solve the technical difficulties encountered by 117 farmers in practice (Khan and Damalas, 2015). Technology demonstrations can provide reference 118 suggestions to farmers in the region for their use decisions. It can effectively eliminate the 119 "worries" of farmers, accelerate radiation and drive neighboring farmers to use biopesticides 120 (Geng et al., 2017). Technology subsidies can effectively reduce the acquisition cost of 121 biopesticides and weaken the uncertain impact of technology and market risks on farmers' income 122 (Gould et al., 2018). 123 However, the theory of technology supply and demand indicates that the real demand of 124 farmers for biopesticides is also the key to determining the success of agricultural extension (Yuan 125 and Niehof, 2011). Previous studies exploring biopesticide adoption behavior from the perspective 126 of agricultural extension note that it can only guarantee and promote the effective supply of 127 biopesticides, ignoring the real demand for biopesticides from farmers. Biopesticide extension 128 should be based on meeting the different technical demands of farmers as the starting point, 129 clarifying the main position of farmers as technology demanders and users, and solving the 130 difficulties encountered in the process of agricultural production and operation as the main goal. 131 For example, the use of biopesticides can protect the ecological environment, increase the 132 profitability of agricultural products, ensure food safety and reduce human health damage (Paul et 133 al., 2020;Constantine et al., 2020). That is, farmers will be motivated to adopt biopesticides when 134 and only when doing so meets their needs and utility goals (Benoît et al., 2020). In this context, 135 the extension of biopesticides and securing market supply will successfully promote the adoption 136 behavior of rice farmers. 137 Based on the above analysis, the following research hypothesis can be obtained: Biopesticide 138 supply and demand will regulate the impact of biopesticide extension on the adoption behavior of 139 rice farmers. Biopesticide extension is more effective when the conditions of "both supply and 140 demand" are met. The framework of this study is shown in Fig. 1. Areas in the Yangtze River Economic Belt" ② was issued in November 2018 and stressed that the 151 Chinese government would strongly support the provinces (cities) in the Yangtze River Economic 152 Belt to implement negative growth in pesticide use. In addition, a number of green pest control 153 demonstration bases should be built to guide farmers to use biopesticides. Hubei Province is an 154 important part of the Yangtze River Economic Belt and one of the main grain production areas in 155 China. Research on its pesticide reduction activity will provide a certain practical reference value 156 for all of central China. More importantly, the Hubei Province Agricultural Management Department made it clear that the "pesticide reduction technology of grain crops" was urgently 158 incorporated into the agricultural technology extension service system ③ . Among these 159 technologies, "biological pest management" is an important technology extension object. 160 Therefore, selecting Hubei Province as the sample area for the study of rice farmers' biopesticide 161 adoption behavior not only has a certain scientific representativeness but also has practical 162 significance for guiding farmers to use more biopesticides. 163 Referring to the list of "Green Control Technology Demonstration Counties" published by the 164 National Ministry of Agriculture and Rural Affairs ④ , our research group randomly selected six 165 counties (districts) from the main rice planting areas in Hubei Province, including the cities of 166 Xiangyang, Huanggang, Jingmen and Yichang (Fig. 2). Combined with the regional distribution of 167 rice production, we selected 6 counties from the above 4 cities (Nanzhang, Wuxue, Qichun, 168 Yingshan, Zhongxiang, Yiling). Then, according to the principle of random stratified sampling, 169 2~3 townships were selected from each county, 4~6 villages were selected from each township, 170 and 10~20 rice growers were selected from each village. Finally, 1148 valid questionnaires of rice 171 farmers in 15 towns and 80 villages were obtained. The questionnaires were completed in 172 face-to-face interviews, with household heads or main agricultural production decision-making 173 members as the respondents. The content design of the questionnaire mainly included detailed 174 data on farmer characteristics, pest control, agricultural production cost, and so on. was constructed using the probit model: 211 where extensioni * represents the latent variable of rice farmers' participation in biopesticide 213 extension, and the value depends on the observable variable extensioni. Vi are other factors 214 affecting farmers' participation in biopesticide extension. Ii is the instrumental variable. λ and κ are 215 coefficients to be estimated. Here, we will draw on Huang et al.'s (2020) study and select "town 216 distance" as the instrumental variable. The more distant the town is, the more difficult it is for rice 217 farmers to access agricultural extension services, and the town distance is a geographic factor that 218 satisfies the exogeneity condition of the model. 219 Second, the model outcome equation (whether to adopt biopesticides or not) was constructed: 220 where biopesticidei * denotes the latent variable of rice farmers' biopesticide adoption, which 223 depends on the observable variable biopesticidei. u is the coefficient to be estimated. δ is a random 224 interference term. Eq. (3) and (4) are fitted to estimate the predicted relationships between the 225 independent and dependent variables in the experimental and control groups, respectively. 226 Finally, a counterfactual scenario is constructed using MILE to obtain the average treatment 227 effect of the impact of biopesticide extension on farmer adoption. For the sample farmers in the 228 experimental group (who already participated in agricultural extension), the difference in the 229 probability of biopesticide adoption between the two scenarios, assuming they were in the 230 nonparticipation scenario, is referred to as "the average treatment effect of the treated (ATT)". 231 Similarly, for sample farmers in the control group (not participating in biopesticide 234 extension), the difference in the probability of biopesticide adoption between them, assuming they 235 were in the participation scenario, is referred to as "the average treatment effect on the untreated 236 In Eqs. (5) and (6), n and m are the sample sizes of the experimental and control groups, 240 respectively. The ATT and ATU values obtained from these calculations can be used as a basis for 241 determining the average treatment effect of biopesticide extension on farmer adoption. In addition, 242 this study will examine the regulatory effect of the "technology supply and demand" of 243 biopesticides using group estimation to verify the heterogeneous impact.  Table 1.   . T-test is the test result of using Stata software to execute "ttest" code to obtain the average value difference between groups. *,**, and*** indicate significant at the 10%, 5%, and 1% significance levels, respectively.

Descriptive statistics 258
We counted the participation of the sample rice farmers in biopesticide extension (Fig. 3). The 259 data showed that the participation of rice farmers in biopesticide extension was 80.99%, 62.91%, 260 43.66%, and 9.86% for the technology publicity, technology demonstration, technology training,

275
Then, we determined statistics on the supply and demand of biopesticides among the sample 276 rice farmers (Fig. 4). It was found that only 14.26% of the farmers were in the effective supply and 277 demand category, and 69.56% of the sample still did not have the demand for adopting

Simultaneous estimation results of the selection equation and result equation 292
Necessary covariance diagnostics were performed for all variables in the model in turn. Then, 293 to identify whether there is an association between the selection equation (whether to participate in 294 biopesticide extension or not) and the outcome equation (whether to adopt biopesticides or not), 295 the two equations were estimated jointly using MILE in Stata14 software, and the results are 296 shown in Table 2. From the correlation coefficients ρ estimated from the selection equation and 297 outcome equation, ρua and ρun passed the significance level test of 10% and 5%, respectively, 298 which shows that the covariance matrix between the equations is indeed correlated. The 299 participation of rice farmers in agricultural extension and the adoption behavior of biopesticides 300 are not completely independent. That is, the use of the ESP model is necessary. 301 302

304
(1) Factors influencing the participation of rice farmers in biopesticide extension 305 From the estimation results of the selection equation, the probability of rice farmers' 306 participation in biopesticide extension was significantly influenced mainly by demand, 307 interviewee age, education, family income, production purpose, scale, and town distance. 308 Specifically, there is no doubt that the supply of agricultural extension services is a prerequisite for (2) Factors influencing the adoption of biopesticides by rice farmers 325 From the estimation of the outcome equation, the adoption of biopesticide by rice farmers 326 was significantly influenced by supply, demand, age, education, family income, production 327 purpose, residue test, brands, selling price, pest cost, environmental cognition, and food safety 328 cognition variables. What is obvious is that technology supply and demand are indeed key factors 329 influencing farmers' biopesticide adoption. Specifically, farmer age negatively influenced the 330 adoption of biopesticides by rice farmers, indicating that the older the rice farmers were, the lower 331 the probability of adopting biopesticides. It is difficult to change the "habits" of older farmers in 332 the short term due to their previous production experience (Huang et al., 2020;Bagheri et al., 333 2021). The direction of influence of education and household income is positive, indicating that 334 highly educated and affluent rice farmers are more likely to adopt biopesticides (Yang et al., 2014). 335 The direction of influence of production purposes is positive because subsistence rice farmers pay 336 more attention to the quality and safety of rice and prefer to use biopesticides with a low toxicity 337 and low residue (Paul et al., 2020). The positive impact of residue tests and brands indicates that 338 both residue tests and brand systems for agricultural products will help increase the probability of 339 active adoption of biopesticides, as the market demand for high-quality agricultural products will 340 push rice farmers to use greener and safer production methods (Li and Guo, 2019).  Notes: a means that this is the "counterfactual" estimation result obtained by MILE. Biopesticide extension in this table refers to any one or more of technical publicity, training, demonstration and subsidy. The number of observations is 1148. *,**, and*** indicate significant at the 10%, 5%, and 1% significance levels, respectively.

364
In addition, the ESP model was used to estimate the effects of four biopesticide extension 365 methods, namely, technology publicity, training, demonstration, and subsidy, on adoption by rice 366 farmers, and the results are shown in Table 3. From the values of ATT and ATU, technology 367 publicity can promote a 7.6% ~ 15.9% increase in the probability of biopesticide adoption by rice 368 farmers. Similarly, the enhancement effect from technology training is 14.8% ~ 15.5%; the 369 enhancement effect from technology demonstration is 17.1% ~ 18.5%; and the enhancement effect 370 from technology subsidies is 5.2% ~ 8.0%. Overall, biopesticide technology demonstration was 371 the most effective way to extend biopesticides, followed by technology training. However, the 372 effect of biopesticide technology publicity and subsidies needs to be improved because their 373 coefficients were small and did not pass the significance test. 374 375

Regulatory effect estimation: Technology supply and demand 376
We then wanted to verify the differences in the effectiveness of biopesticide extension under 377 different biopesticide supply and demand scenarios. The samples were divided into two 378 subsamples three times according to biopesticide supply and demand. The impact of biopesticide 379 extension on farmers' adoption behavior in different subsamples was estimated again using the 380 ESP model, and the results are shown in Table 4. First, for the "only demand" samples, the mean 381 treatment effects ATT and ATU of biopesticide extension on adoption by rice farmers were 0.078 382 and 0.066, indicating that agricultural extension increased the probability of biopesticide adoption 383 by 6.6% ~ 7.8%. Second, for the "only supply" samples, the mean treatment effects ATT and ATU 384 of biopesticide extension on adoption by rice farmers were 0.105 and 0.096, indicating that 385 agricultural extension increased the probability of biopesticide adoption by rice farmers by 9.6% ~ 386 10.5%. Third, for the "both supply and demand" samples, the mean treatment effects ATT and 387 ATU of biopesticide extension on rice farmers' adoption were 0.361 and 0.247, indicating that 388 agricultural extension increased the probability of biopesticide adoption by rice farmers by 24.7% 389 ~ 36.1%. Fourth, for the subsamples of "no supply", "no demand" and "no supply or demand", all 390 results estimated in the group failed the significance test. 391 The above results can be connected to two important findings. On the one hand, the supply 392 and demand of biopesticides significantly regulated the effect of biopesticide extension on the 393 adoption behavior of rice farmers. In the scenario with no effective supply and demand for 394 biopesticides, agricultural extension is very ineffective, which is an important element to explain 395 the poor effectiveness of biopesticide extension in China. The research hypothesis was verified. 396 On the other hand, improving the supply-demand equilibrium of biopesticides can substantially 397 improve the effectiveness of biopesticide extension. The estimated results for the "both supply and 398 demand" sample group are 2.35~5.47 times higher than those for the "only supply" and "only 399 demand" groups. Therefore, increasing the effective supply and demand of biopesticides is the top 400 priority to promote the rapid expansion of biopesticides in China. 401 402 Notes: Other result information is omitted here, and only ATT and ATU values are shown. *,**, and***indicate significant at the 10%, 5%, and 1% significance levels, respectively.

404
The above empirical evidence has found that the effective supply and demand of 405 biopesticides is the key to determining the effectiveness of agricultural extension. However, 406 differences in the roles of different biopesticide extension approaches need to be further explored. 407 The sample farmers were divided into two groups: "both supply and demand" and "no supply or 408 demand". Then, the ESP model was used to estimate the average treatment effects ATT and ATU 409 for the effects of technology publicity, training, demonstration and subsidies on the probability of 410 farmers' adoption. What is very clear from the results in Table 5 is the estimated difference 411 between groups. First, in the "both supply and demand" scenario, the effect of any type of 412 biopesticide extension is always significant. This again validates the regulatory effect of "supply 413 and demand". Second, distinguishing the results in Table 3, effective supply and demand led to a 414 substantial increase in the effectiveness of technology publicity, training, demonstration, and 415 subsidies for biopesticide extension. Thus, the research hypothesis was reaffirmed. 416 417 Notes: Biopesticide extension in this table refers to any one or more of technical publicity, training, demonstration and subsidy. *,**, and***indicate significant at the 10%, 5%, and 1% significance levels, respectively.

Heterogeneous impact on farmers with different characteristics 421
In addition, we need to account for the impact of sample heterogeneity on the study findings. 422 The sample was grouped in terms of the characteristics of scale, age, education, and 423 co-organization, and the results in Table 6 were obtained after sequential estimation. Obviously, 424 regardless of the characteristics of farmers, "both supply and demand" is still a strong guarantee of 425 the effectiveness of biopesticide extension. Moreover, the results showed that biopesticide 426 extension had a greater impact on smaller-scale, younger, less educated, and cooperative member 427 sample farmers, whose average treatment effects on the probability of biopesticide adoption were 428 29.6%~30.1%, 20.5%~23.1%, 25.4%~30.4%, and 24.7%~25.5%, respectively.

451
Unlike previous scholarly studies arguing for a causal relationship between agricultural extension 452 and farmers' biopesticide adoption (Toepfer et al., 2020), this study further compared and 453 contrasted the impact effects of four specific agricultural extension approaches: technology 454 publicity, training, demonstration, and subsidies. Among them, we found that biopesticide 455 technology demonstration is the most effective extension method. There is still a need to increase 456 the construction of a biopesticide demonstration base so that farmers can see the "seeing is 457 believing" effect for pest management (Guo and Wang, 2016). Of course, the most important 458 finding is that insufficient supply and demand for biopesticide products are the core factors 459 leading to low agricultural extension effectiveness. In our study, we found that 85.74% of the 460 sample farmers affected the effective supply demand for biopesticides due to low demand or 461 insufficient supply, which will be a major problem to be solved in the future extension of 462 biopesticide technology publicity, training, demonstration and subsidies. We must also insist on 507 strengthening technology demonstration work to reduce farmers' worries about using biopesticides. 508 Of course, to consider the equity and efficiency of agricultural extension issues, a wider range of 509 farmers can participate in the promotion of biopesticides. On the other hand, it is important to 510 focus on the excellent attributes of biopesticide products, accurately identify and guide 511 agricultural producers' demand for biopesticides, and then make farmers more active and positive 512 in using biopesticides. At the same time, it is also necessary to increase the product market supply 513 of biopesticides. More importantly, we should focus on smaller, younger, less educated and 514 cooperative member farmers to carry out biopesticide extension to obtain a greater marginal 515