• Geo-networks positively impact farmers' land transfer behavior

According to the regression results of the Probit model, the number of farmers making land transfer in the same village (X1) was positively significant at 1% with a coefficient of 0.358, showing increasing marginal effects. For each unit increase in the number of farmers in the same village making land transfers, farmers' land transfer behavior was 1.102 times its original value, demonstrating that the number of farmers making land transfer in the same village positively affected farmers' land transfer behavior. The reason for this is that farmers in the same village are in the same geo-networks that a farmer typically interacts with most frequently, and they are crucial in transmitting information about land transfer. Individual farmers may also imitate the land transfer behavior of other farmers in the same village, because these individuals, uncertain about their judgments, tend to follow the majority. Hence, the probability of farmers making land transfers increases as more land transfer occurs in the village, affirming the herd effect in farmers' land transfer behavior.

The number of village cadres in the same village making land transfers (X2) was positively significant at 1% with a coefficient of 1.649, presenting increasing marginal effects. For each unit increase in the number of village cadres making land transfers, farmers' land transfer behavior was 1.129 times its original value. This means that the number of village cadres making land transfers positively affected the land transfer behavior of farmers in the same village. This is because the village cadres are the organizers and leaders of the villagers in their respective villages, and they access more information about land transfer policies and information, hence taking the role of releasing and disseminating information. Village cadres are highly respected by farmers and provide them with support; their decisions often direct farmers' actions. The number of farmers in the same village making land transfers and the number of village cadres in the same village making land transfers, as two geo-network variables, positively influenced farmers' land transfer behavior. Hence, hypothesis H1 is verified.

• Impact of control variables on farmers' land transfer behavior

The age and gender of the householder, the area of arable land operated by farmers, and farmers' satisfaction with farmland infrastructure were all related to farmers' land transfer behavior. Specifically, the age of the householder (X3) was negatively significant at 10% with a coefficient of −0.015, showing decreasing marginal effects. This denotes that the younger the householder, the higher is the probability of land transfer. The gender of the householder (X4) was negatively significant at 5% with a coefficient of −0.502, showing decreasing marginal effects. This indicates that male householders are more likely than female individuals to transfer their land.

The arable land area (X7) was positively significant at 1%, with a coefficient of 0.541, and farmers' land transfer behavior was raised 2.603 times its original value for each unit increase in arable land area, showing increasing marginal effects. This may be attributed to the fact that increased cultivated land areas require a longer operating cycle, and more economic inputs lead to higher earnings. Farmers continue transferring land inward to enlarge their farming scale for financial gain, hence the probability of land inward transfer grows. Meanwhile, as the arable land area continues to enlarge, economic inputs are positively proportional to the risks facing cultivated land. In other words, the more economic inputs, the greater are the risks involved, hence the increased possibility of outward land transfer.

Farmers' satisfaction with farmland infrastructure (X14) is positively significant at 5%, with a coefficient of 0.189, suggesting increasing marginal effects. Increased satisfaction among farmers with farmland infrastructure was associated with greater probability of land transfer. This indicates that farmers' satisfaction with farmland infrastructure positively affects farmers' land transfer behavior. The better the farmland infrastructure, the more favorable it is for agricultural production. Farmers conducting land transfer tend to have better farmland infrastructure and thus earn more rent, and those who transfer land inward can benefit from upgraded farmland infrastructure, which will raise agricultural output and allow additional land transfer activities.

#### 5.1.3. Verification of Herd Effect in Farmers' Land Transfer Behavior

Probit regression cannot effectively address the correlation effect and reflexivity when identifying the herd effect [18]. In order to overcome possible endogeneity in the Probit model, this study developed the IV-Probit model for regression analysis of the sample data and tested the validity of the instrumental variable of the area where farmers are located (IV) using weak instruments. The regression results are tabulated in Table 4. The first-stage F statistic of the IV-Probit was 11.43, greater than the empirical value of 10. The weak identification shows that the *p*-values of the Anderson–Rubin and Wald tests are positively significant at 5%, demonstrating that the instrumental variable selected in this paper was not a weak instrument. This proves that farmers' land transfer behavior imitates the behavior of those in the same group within a geo-network, and the herd effect exists. Hence, hypothesis H2 is verified.


**Table 4.** Herd effect in farmers' land transfer behavior: IV Results.

Note: \*\*\* denotes positive significance at 1%.
