5.1. Validation of Model
Before the structural equation model is used for analysis, reliability and validity tests should be carried out on the data. Table 2
includes standardized factor loadings, Cronbach’s α, AVE, and CCR for each construct. All the loadings (see Figure A1
) of items that remained to measure each variable were above 0.7 and larger than the cross-loadings, all items for each construct were confirmed to satisfy the criteria. Cronbach’s α is an important index to check the internal consistency of latent variables, the Cronbach’s α values for motive, opportunity, operation, anti-risk, trust, and willingness were 0.836, 0.863, 0.762, 0.776, 0.759, and 0.852, respectively; each construct was above 0.70. The AVE values were between 0.5622 and 0.7431, and all above 0.50, indicating a high level of convergent validity [27
]; and the CCR were between 0.764 and 0.870, surpassing the suggested 0.70 minimum and indicating that the composite measurement items have sufficient internal consistency reliability [27
]. In addition, the KMO Test value was 0.835 and the Bartlett’s Test statistical value was 0.000, indicating that the observed variable data had good validity and was suitable for factor analysis. Therefore, those tests show that this measurement model has a good convergent validity.
Before the initial model to be modified, cross-validation should be conducted [28
]. Therefore, we used SPSS20.0 (IBM, New York, NY, USA) to randomly divide the original sample into two parts: a derivation sample and a cross validation sample. Then, two groups were established in AMOS, and the randomly divided sample data were imported to conduct multi-group analysis to obtain cross-validation results. The results (see Table 3
) show that the hypothesis “Assuming model unconstrained to be correct” is accepted (p
> 0.1). It is suggested that once a favored model is found, its fit can be assessed by using different data from the cross-validation sample. The cross-validation test indicates that the validity of the model is ductile and stable, which further demonstrates the validity of the model.
5.2. The Results of SEM
AMOS21.0 software (IBM, New York, NY, USA) was used to construct a structural equation model of the relationship among adoption motivation, adoption opportunity, technical operational ability, anti-risk ability, trust and adoption willingness. Before presenting the results for the structural equation model, we will first present the fitness test for the justification for the SEM. In this paper, absolute fit index, incremental fit index and parsimonious fit index are considered when evaluating the fitness of the model, so that a consensus can be reached on the acceptability or rejection of the model. Since AGFI is lower than 0.9 and CMIN/DF is higher than 3 in the initial model, the index was modified by adding a correlation of error terms, and obtaining the fitness index of modified model (see Table 4
). The absolute fit index, incremental fit index, and parsimonious fit index of the modified model are all within the adapted standards [30
], suggesting that the actual data fit well with the theoretical model constructed above.
According to the data analysis in Table 5
and the results of the structural equation model in Figure 2
and Figure 3
, the following results were obtained:
(1) The adoption motivation has a significant positive impact on the operational ability of farmers, and also has a significant positive impact on the adoption willingness. The results showed that both H1a and H1b were verified. It also shows that the adoption motivation influences the adoption willingness mainly through two paths: The first path is “motivation → willingness”, that is, adoption motivation directly affects the adoption willingness, indicating that the stronger the adoption motivation of farmers, the higher their adoption willingness will be. The second path is “ motivation → operation → willingness”, that is, the adoption motivation is indirectly affected the adoption willingness through the technical operation ability, which confirms that the stronger the adoption motivation of farmers is, the stronger the operation ability is, thus contributing to the improvement of adoption willingness.
(2) The influence of adoption opportunity on adoption willingness, adoption motivation, and trust is significantly positive, among which adoption opportunity has the greatest influence on trust. The research hypotheses H2a, H2b, and H2e have been verified. It can be seen from the results that the adoption opportunity mainly influences the adoption willingness in two ways: one is “opportunity → willingness”, that is, the adoption opportunity directly affects the adoption willingness, indicating that the more adoption opportunities farmers have, the stronger their adoption willingness will be. The second path is “opportunity → motivation → willingness”, that is, adoption opportunity indirectly affected adoption willingness through adoption motivation, which proves that the more adoption opportunities farmers have, and the stronger their adoption motivation will be, thus contributing to the improvement of adoption willingness. Therefore, with the increase of adoption opportunities, farmers’ willingness to adopt green fertilization technology was significantly enhanced. At present, the government often improves the chances of adopting green fertilization technology by providing information, technology, and purchasing services for farmers. However, the survey results showed that among the sample farmers, the proportion of farmers who agreed that the government played a significant role in information, technology, and purchase of services was relatively low, which indicated that improving the socialized service ability of the government could greatly improve the chances for farmers to adopt green fertilization technology.
(3) Both technical operational ability and anti-risk ability have a significant positive impact on the adoption willingness, and the technical operational ability also has a significant positive impact on the adoption willingness of farmers by acting on the ability to resist risks. This indicates that improving the operational ability and anti-risk ability of farmers will help increase the willingness of farmers to adopt the green fertilization technology. The hypotheses H3a, H3b and H3c have been verified. It can be seen that reducing the risks in the process of adopting green fertilization technology and improving the technical operation ability of farmers will greatly enhance their willingness to adopt green fertilization technology. At present, the technical operation ability and anti-risk ability of farmers are relatively low, mainly because the application threshold of green fertilization technology is relatively high, which is difficult for farmers to master. In addition, the low household income level and imperfect supporting measures will increase the risk of green fertilization technology application, and the low knowledge level of most farmers will also hinder the improvement of farmers’ technical operation ability. Therefore, even though government technical training and agricultural subsidies, the improvement of farmers’ technical operation ability and risk resistance ability is still limited, it is difficult for the smallholders to adopt.
(4) Trust did not have a significant direct impact on the adoption willingness, and H4a was not verified. However, trust indirectly positively affected the adoption intention through operational ability, with a path coefficient of 0.285, which indicated that enhancing the degree of trust in the government of farmers would help indirectly improve the willingness of farmers to adopt green fertilization technology. The research hypothesis H4c proposed above has been verified. Therefore, the spread and application of green fertilization technology in China cannot be promoted without farmers’ high trust in the government.
The total effect is the sum of the regression coefficients of all direct effects and indirect effects related to the latent variable in the path model. The strength of indirect effects is multiplied by the standard regression coefficients of direct effects between the two endpoint variables. In this model, the total effects of green fertilization technology adoption motivation, adoption opportunity, technical operation ability, anti-risk ability and trust on the willingness to adopt were 0.610, 0.381, 0.491, 0.297 and 0.259, respectively. It can be seen that the adoption motivation is the most important factor affecting the willingness of farmers to adopt green fertilization technology, followed by technical operation ability and adoption opportunity.
Existing research shows that cooperative participation and production and sales contract signing are beneficial for farmers to adopt green fertilization technology [31
]. Therefore, a cross-group structural equation model was used in this paper to explore the willingness and driving path of different types of farmers to adopt green fertilization technology. The results of the cross-group structural equation model are shown in Table 6
, and the indirect influence path is omitted. Except that the fitness index AGFI is less than 0.9, but very close to 0.9, the test indexes of the adjusted model all meet the requirements, and the model is generally capable of interpretation. According to the direct effects in the table, organizational participation and contract signing can regulate the positive effects of adoption opportunity, adoption motivation, technical operation ability and risk resistance on the willingness to adopt green fertilization technologies. In the group of farmers who participated in the organization or signed the contract, the adoption opportunity and technical operational capability had a significant positive impact on the adoption intention, but this path was not significant in the group of farmers who did not participate in the organization or signed the contract. Conversely, in non-participating or non-contracted farmer groups, anti-risk capabilities have a significant positive impact on adoption willingness, but the path is not significant in participating organizations or contracted farmer groups. In addition, adoption motivation had a significant positive effect on adoption intention in all the farmer groups.