3.3.3. Control Variable

In addition to household structure, other variables of individual characteristics and household characteristics also affect land transfer. Variables of individual characteristics include gender, age, education, ethnicity, and health status of the householder, while variables of household characteristics include the logarithm of household income, household land area, labor ratio, and participation in professional cooperatives. Household characteristics also affect land transfer. Variables of individual characteristics include gender, age, education, and ethnicity.

Table 1 shows the mean statistics of the variables used in this paper. Column (1) is the mean statistics of the full sample; column (2) is the mean statistics of the variables related to the elite household; column (3) is the mean statistics of the variables related to non-elite households. In the full sample, the mean of the area of land subcontracted to farmers for compensation is 0.563 mu, while for elite households, the mean of the area of land subcontracted to farmers for compensation is 0.762 mu. Compared with 0.523 mu of non-elite households, it is 0.239 Mu higher. The descriptive results also show that the percentage of elite households in the total sample is 16.6% and the percentage of incomplete households is 8.2%. The mean of other variables can be accessed from Table 1 and is not repeated here.


#### **Table 1.** Descriptive statistics of variables.

#### 3.3.4. Model Building

The subject of this paper is the effect of family structure heterogeneity on land transfer area to investigate whether elite households differ from incomplete households in land transfer. With the help of existing studies, family structure is set as the main explanatory variable, land transfer area as the explained variable, and individual characteristics and household characteristics as control variables. According to the research hypotheses, the following models are set-up in this paper.

$$\text{Land outi} = \alpha + \beta \text{1FM\\_st1} + \gamma \text{i} \chi \text{i} + \varepsilon \text{i} \tag{1}$$

$$\text{Land outi} = \alpha + \beta \text{1FM\\_st0} + \gamma \text{i\chi} + \varepsilon \text{i} \tag{2}$$

$$\text{Land outi} = \alpha + \beta \,\text{1 FM\\_st1} + \beta \,\text{2FM\\_st0} + \gamma \text{i}\chi\text{i} + \varepsilon \,\text{i} \tag{3}$$

The above-explained model variables are all land area subcontracted to individuals for a fee, and the main explanatory variable in model (1) is elite households; the main explanatory variable in model (2) is incomplete households; model (3) is a full-variance model, and, at the same time, the two variables of elite family and whether it is an incomplete family are added. Land out in models (1)–(3) means land area subcontracted to individuals for compensation. FM\_\_strl in the above model is a proxy for elite households, FM\_ str0 is a proxy for incomplete households, and χi are control variables. In the empirical analysis

and robustness test sections, the independent variable measures are replaced in order to analyze the influence mechanism and test the robustness of the results.
