2.3.3. Difference-in-Differences Model

The difference-in-differences model [45] involves comparing the effect of a research subject before and after the intervention of a specific factor, and the difference between the two is the net effect of that factor on the research subject [34]. The basic idea is to divide the survey sample into two groups: one group of subjects affected by the specific factor, namely the "experimental group", and one group of subjects not affected by the specific factor, namely the "control group". The specific model is expressed as follows:

$$Y\_{\rm ij} = \beta\_0 + \beta\_1 Treated\_{\rm ij} + \beta\_2 Period\_{\rm ij} + \beta\_3 DID\_{\rm ij} + \varepsilon\_{\rm ij} \tag{6}$$

$$DID\_{ij} = Treatment\_{ij} \* Period\_{ij} \tag{7}$$

where *i* = 1 represents the pre-intervention period, *i* = 2 represents the post-intervention period, *j* represents the subject, *Yij* represents the value to be measured for the jth subject in period I, *Periodij* is a time dummy variable, *Period1j* = 0 represents pre-intervention, *Period2j* = 1 represents post-intervention, *Treatij* is a group dummy variable, *Treatij* = 0 is the control group, *Treatij* = 1 is the experimental group, *DID* is the cross term of *Treatij* and *Periodij*, the *εij* is the unobserved other variables affecting *Yij* controlled not to change.

Depending on the characteristics, it is possible to write separate models of changes in the variables to be measured in the control and treatment test subjects, before and after the factor intervention.

The control group *Treatij* = 0 was modeled as *Yij* = *β*<sup>0</sup> + *β*2*Periodij* + *εij*. Therefore, the values to be measured for the control group in the periods before and after the factor intervention were:

$$\chi\_{ij} = \begin{cases} \beta\_0 + \varepsilon\_{ij\prime} & i = 1 \\ \beta\_0 + \beta\_2 + \varepsilon\_{ij\prime} & i = 2 \end{cases} \tag{8}$$

The changes in the values to be measured in the control group before and after the factor-specific intervention were:

$$diff\_1 = \left(\beta\_0 + \beta\_2 + \varepsilon\_{i\bar{j}}\right) - \left(\beta\_0 + \varepsilon\_{i\bar{j}}\right) = \beta\_2 \tag{9}$$

The experimental group *Treatij* = 1 was modeled as *Yij* = *β*<sup>0</sup> + *β*<sup>1</sup> + *β*2*Periodij* + *β*3*Periodij* + *εij*. Therefore, the values to be measured for the control group in the two periods before and after the factor intervention were:

$$Y\_{\rm ij} = \begin{cases} \beta\_0 + \beta\_1 + \varepsilon\_{\rm ij}, & i = 1 \\ \beta\_0 + \beta\_1 + \beta\_2 + \beta\_3 + \varepsilon\_{\rm ij}, & i = 2 \end{cases} \tag{10}$$

The changes in livelihood capital in the experimental group before and after the ad hoc factor intervention were:

$$diff\_2 = \left(\beta\_0 + \beta\_1 + \beta\_2 + \beta\_3 + \varepsilon\_{ij}\right) - \left(\beta\_0 + \beta\_1 + \varepsilon\_{ij}\right) = \beta\_2 + \beta\_3\tag{11}$$

Thus, the net effect of a given factor on the observations of the subject to be measured is:

$$diff = (\beta\_2 + \beta\_3) - \beta\_2 = \beta\_3 \tag{12}$$

The final value *β<sup>3</sup>* is the final double difference value to be obtained. When *β<sup>3</sup>* > 0, it indicates that the specific factor had a positive effect on the study subject; when *β<sup>3</sup>* < 0, it indicates that the specific factor's effect was negative effect. The larger the absolute value of *β3*, the greater the degree of influence of the specific factor on the study subject.

### **3. Results and Analysis**

#### *3.1. Analysis of Changes in Livelihood Capital of Different Types of Farmers*

#### 3.1.1. Description of Differences in Livelihood Capital of Different Types of Farmers

As shown in Table 6, the livelihood capital of non-agricultural and non-agricultural part-time farmers before land transfer was higher, with values of 2.553 and 2.309, respectively, while that of purely agricultural and agricultural part-time farmers was lower, with respective values of 2.039 and 2.241. This indicates that part-time farming has a positive effect on the livelihood capital of farmers.

Specifically, natural capital was highest for purely agricultural farmers, followed by agricultural part-time farmers and non-agricultural part-time farmers, while nonagricultural farmers had the lowest natural capital index values; the values were 0.459, 0.401, 0.374, and 0.358, respectively. The natural capital of these farmers was lower because they were engaged in non-agricultural activities, as non-agricultural farmers are mainly engaged in non-agricultural activities to maintain their livelihoods.

Financial capital was highest for non-agricultural farmers, followed by non-agricultural part-time farmers and agricultural part-time farmers, and the lowest values were for purely agricultural farmers, with indicator values of 0.573, 0.392, 0.376, and 0.297, respectively. Differences in financial capital of farmers were found in terms of annual household income, and farmers engaged in non-agricultural activities had a wider range of livelihood sources. Most of them go out to work, so their income is more stable than farming and is not limited by land quality and natural conditions.


**Table 6.** Values of livelihood capital indicators before and after land transfer for different types of farmers.

Livelihood capital was highest for non-agricultural farmers, followed by non-agricultural part-time farmers and agricultural part-time farmers, and the lowest livelihood capital was for purely agricultural farmers, with livelihood capital values of 0.372, 0.361, 0.342, and 0.304, respectively. Differences in livelihood capital were primarily found in the two indicators of housing quality and availability of transportation, with non-agricultural and part-time farmers not simply dependent on the land for their livelihood, but having a wider variety of livelihood sources and higher living capital.

Purely agricultural and agricultural part-time farmers had greater levels of productive capital than non-agricultural part-time farmers, who had the lowest levels. Production capital returned the following values: 0.335, 0.324, 0.331, and 0.277, correspondingly. The differences in productive capital were primarily due to differences in the number of productive tools, with farmers who were primarily dependent on land as a source of income generally acquiring more productive tools. Additionally, local government is strengthening the construction of farmland water conservation to increase production.

Human capital was highest for non-agricultural farmers, followed by part-time farmers, and was lowest for purely agricultural farmers, with indicator values of 0.546, 0.494, 0.487, and 0.373, respectively. Differences in human capital were mainly manifested in the educational level and health status of the labor force. Non-agricultural farmers had relatively higher education levels and filled a wider range of occupations.

Social capital was highest for non-agricultural farmers, followed by part-time farmers, and lowest for purely agricultural farmers, with indicator values of 0.427, 0.357, 0.311, and 0.271, respectively. As farmers increase their part-time employment, their understanding of national policies increases, and the social security they can enjoy also increases. In the process of engaging in more part-time employment, farmers' social interactions increase and their social capital increases.

3.1.2. Analysis of the Direction of Change in Livelihood Capital of Different Types of Farmers

Combining Table 6 with Figure 2 above, it can be calculated that the capital worth of pure farmers' livelihoods after land transfer was 2.177, which was 0.138 more than before the land transfer. Agricultural part-time farmers' livelihood capital value was 2.346, which was 0.105 higher than before circulation. Non-agricultural part-time farmers' livelihood capital value was 2.390, which was 0.081 higher than before circulation. Non-agricultural farmers' livelihood capital value was 2.578, which was 0.025 higher than before circulation. When compared to other capital, pure farmers' productive capital expanded greatly, while natural and financial capital increased only somewhat. Pure farmers tended to participate in land transfer by moving out plots of land that are far away or of poor quality, keeping plots of land that are of superior quality, and moving onto existing land near to their land to increase their operational scale. To boost the productivity of the remaining land, better instruments were obtained for it the same time.

**Figure 2.** Direction of change in livelihood capital of different types of farmers.

In contrast to pure farmers, agricultural and non-agricultural part-time farmers' natural capital declined while their social capital rose dramatically. The other components of their capital were virtually unaffected. Compared to non-agricultural part-time farmers, who continued to focus primarily on land management, they experienced less loss of natural capital. However, these farmers work part-time jobs that will improve their social interactions, diversify their sources of income, and raise their incomes, so building their social and financial capital.

Natural and productive capital in non-agricultural farmers declined, whereas livelihood capital increased. Human, financial, and social capital all changed less, and livelihood capital did not change considerably. Because non-agricultural farmers have forgone agricultural activities and have an income that is entirely unrelated to agriculture, the process of land transfer essentially has no impact on these farmers' ability to support themselves. As a result, the change in their livelihood capital was insignificant.
