Promoting Sustainability: Land Transfer and Income Inequality Among Farm Households

Abstract

The United Nations Sustainable Development Goals emphasize the dangers of inequality and initiatives on how to reduce it. Income inequality is an important part of this and can cause many social problems. This study aims to investigate the impact of land transfer on income inequality of Chinese farm households and its mechanism of action. This study empirically analyzes the impact of land transfer on farm household income inequality based on the OLS model using 27,134 samples from the 2018–2020 China Family Panel Study (CFPS) data. The study finds that land transfer can effectively reduce income inequality among farm households. However, only land transfers out affect income inequality; the effect of land transfers in is not significant. At the same time, there is some heterogeneity in the impact of land transfers on farm household income inequality. The effect is greater in the east and west than in the center. The effect is greater in the north than in the south. The role is greater in food-producing areas than in non-food-producing areas. Mechanism testing shows that land transfer impacts income inequality among farm households by promoting entrepreneurship. Farm households who use the Internet and have stronger social capital gain more from the impact of land transfer on income inequality.

1. Introduction and Literature Review

Over 30 years, more than 1 billion people have been lifted from extreme poverty. The world has come a long way in reducing poverty. However, during this period, the income share of the poorer half of the population has remained virtually unchanged even though global economic output has more than tripled since 1990. Inequality undermines economic progress and, in turn, exacerbates the social divides that inequality creates. In China, the world’s largest developing country, inequality also exists. China’s rapid economic development since its reform and opening up has created a “growth miracle”, but scholars recognize that the gap between rich and poor in China is also growing. A study showed that China’s Gini coefficient had reached between 0.53 and 0.55 around 2010 and that the main reasons for this income inequality were regional differences and the urban–rural gap [1]. Income inequality can have a severe impact on the stability of a society and take a toll on people’s mental and physical health. Therefore, research on income inequality is necessary.

Income inequality is a hot topic for economists and sociologists. Income inequality is the unequal income distribution among individuals or households in a given society, economy, or group [2]. Income inequality is not specific to a particular country but is a universal phenomenon. It is worth noting that income inequality is not limited to countries but can also exist within a country. A gap exists between developing countries’ rural and urban living standards, measured in income, consumption, or non-monetary aspects. Statistical data for 108 countries also show that, in today’s world, the income gap between countries has narrowed significantly, while the gap between regions within countries has widened [3].

An early researcher on income inequality was Kuznets, who proposed the “Kuznets Curve”, which suggests that income inequality increases and then decreases as the economy develops, showing an inverted “U-shape” [4]. The Kuznets curve concept has been widely used in determining income inequality and has been validated in many countries. Researchers have found that the phenomenon may be present in North America, Europe, and a wide range of developing countries, including China [5]. They have argued that China has passed the turning point of the Kuznets Curve and that income inequality is on a downward trend. Recently, however, some scholars have argued differently. Using new inequality indicators, they found that the Kuznets growth model has little explanatory power for income inequality in China. Land policies have likely played a key role [6]. However, whether or not they agree with the applicability of the Kuznets curve in China, their conclusion that the deterioration of income distribution in China has been reversed is unanimous [7].

As research has progressed, scholars have realized that income inequality may also affect economic growth [8]. In the 1990s, several Southeast Asian and South American countries entered the middle-income stage, but economic growth stagnated along with it. Scholars believe this was partly because these countries were not innovative enough to compete with developed countries in capital- and technology-intensive industries and lost many markets. On the other hand, growing income inequality in these countries has affected social stability, reduced investment, and hindered economic growth. This result is confirmed in panel data for 71 countries [9]. As the 21st century progressed, income inequality was found to play a different role in different income countries [10]. Barro argued that the relationship between income inequality and economic growth might not be linear and that income inequality could dampen economic growth for countries with a GDP per capita of less than USD 2070 [11]. Above that value, however, income inequality promotes economic growth. Income inequality may lead to other negative consequences as well. One key discussion is population health. It has been argued that income inequality worsens the health of populations and increases mortality [12]. This phenomenon exists in Europe, the United States, and Japan [13]. Social cohesion may be an important mechanism of action [14]. However, some scholars, using data from Canada and the United States, have found that the phenomenon occurs only in the United States, possibly because of differences in distribution between the two countries [15]. In addition, income impossibilities may also lead to unequal consumption and increased carbon emissions [16,17], deteriorating environmental quality [18].

For the study of income inequality in China, the perspective mainly focuses on urban–rural areas. The data show that China’s urban–rural income ratio was 2.39 in 2023, much higher than the average of OECD countries. Researchers have found that urban–rural income inequality in China is the leading cause of income inequality. In addition, factors such as population aging [19], international trade [20], tourism, and the development of the traditional financial sector may also exacerbate income inequality in China [21,22]. Moreover, due to China’s dualistic economic structure, the impact of the same variable may be different and heterogeneous in other regions. East, center, west, south, north, big, and small cities likely exhibit different characteristics. This will make it difficult for a single policy to play a decisive role in income inequality. In recent years, with the advent of the digital age, information and communication technologies (ICTs) have had a new impact on income inequality. Researchers have found that using the Internet reduces income inequality among French workers [23]. The same evidence emerges for China’s young labor market, and the returns to long-term Internet use are higher than those to short-term Internet use [24].

Most of the previous studies on income inequality are macro-level [25,26,27]. Our study is more micro, placing income inequality within rural areas. The rationale for this is that the urban–rural income gap in China has received much attention, and the Chinese government has implemented many policies to try to narrow it. It is also true that the urban–rural income gap in China has been declining in recent years. However, the income gap within rural China has not been emphasized. According to China’s National Bureau of Statistics, the rural per capita income ratio between the regions with the highest and lowest rural per capita incomes in 2022, Shanghai and Gansu, was 3.26, much higher than China’s urban–rural income ratio. More importantly, despite regional differences, income inequality in the same village can be caused by differences in farmers’ land, labor, capital, technology, and other factors [28]. Therefore, studying income inequality based on this perspective within the farm household is significant.

Land transfer is the process of transferring land use rights out of or into a rural area. Land transfer is usually a two-way street, either transferring one’s land to another farm household or economic organization, or obtaining land from another farm household or financial organization. Transferring one’s land to another farmer or economic organization can be referred to as land transfer out, and acquiring land from another farmer or financial organization can be referred to as land transfer in. However, it should be noted that whether transferring out or transferring in, what is acquired is only the right to use the land, not the right of ownership. This phenomenon in China is very different from that in other countries. China’s large-scale land transfer began in 2008, and the area of land transfer has rapidly expanded from 109 million mu in 2008 to 512 million mu in 2017 (the year-by-year figures are 109, 150, 187, 228, 341, 403, 447, 479, and 512 million mu). The reason behind such a large-scale land transfer is that cities need more labor. Farm households who transfer their land will be more likely to work in cities, speeding up urbanization. Land transfer is believed to increase the efficiency of land use, enhance large-scale operations, and reduce the use of agricultural fertilizers, further reducing environmental pollution [29]. Land transfer also reduces agricultural land abandonment due to labor costs and increases farm households’ income [28,30].

Research on land transfer and farm households’ income inequality has been discussed less in the academic community. Although most studies agree that land transfer can increase farm households’ income [31,32], it does not mean it can necessarily reduce income inequality. This is because there is a herd effect in the land transfer behavior of farm households [33], and the land transfer behavior of farm households is not always appropriate. Previous studies have also shown that farm households will have differences in the benefits they receive from land transfer because of some of their characteristics [34]. The benefits are different in different areas [35]. Higher-educated farm households who skillfully master new technologies are likelier to have high returns [36]. Therefore, there is a need to conduct an in-depth study on the impact of land transfer on income inequality among farm households [37].

The research objective of this study is to examine land transfer’s impact on income inequality among farm households. It can contribute to the existing literature in the following four ways. First, it expands the perspective of research on income inequality and increases the academic attention to income differences within farm households. Second, it enriches the impact of land transfer on farm household income and clarifies that land transfer can also reduce income inequality of farm households on top of its impact on increasing farm household income. Additionally, the effect of land transfer out is greater than that of land transfer in. Thirdly, it puts forward the role of the channel of land transfer affecting farm household income inequality and tries to solve the problem of endogeneity, which provides help for academics in the causal identification between the two variables and theory construction. Fourth, it provides a case study in developing countries to study income inequality.

2. Theoretical Analysis and Research Hypothesis

For farm households, land is the primary means of earning a living and often the main means of accumulating and transmitting wealth from one generation to the next [38]. Therefore, land rights changes are of particular importance to farm households. According to the concept of livelihood diversification proposed by Ellis, restricted land markets can hinder the process of livelihood diversification for farm households [39]. Therefore, the impact of land transfers on income inequality among farm households is likely to occur mainly through the promotion of livelihood diversification. On the one hand, land transfers can change the way farm households allocate their labor and increase livelihood diversification. For example, farm households involved in land transfer will be more willing to migrate to the city [40] and more inclined to get jobs in the non-farm sector, accelerating the transfer of surplus agricultural labor at the macro-level. In this process, factors of production such as labor are allocated from the agricultural sector to the non-agricultural sector, so the non-agricultural income of farm households will increase, and livelihood diversity will be enhanced. Coupled with the fact that low- and middle-income farm households are more dependent on land and have a smaller income base than high-income farm households, the increase in the income of low- and middle-income farm households involved in land transfer will be more pronounced. On the other hand, even if farm households involved in land transfer do not choose to move to the cities, they can still engage in other local production activities; for example, they can invest the capital gained from land transfer in farming, various entrepreneurial activities, or even part-time jobs. This group of farm households is transitioning from low-value-added crop cultivation to higher-value-added livestock, aquaculture, fisheries, entrepreneurship, and other livelihood activities to increase the efficiency of agricultural production [41]. In the process, they have increased their sources of income and diversified their livelihoods. However, it is essential to note that land transfers in may be less effective than land transfers out. Theoretically, the participation of farm households in land transfers is bound to help maximize net incomes by increasing their incomes. Still, the challenges faced by different transfers are different in reality. Generally speaking, the outcome of land transfers out is not very different from what was expected at the time of decision-making. In contrast, land transfers are subject to several uncertainties (e.g., natural disasters and planning adjustments) that go beyond what was expected. This may be such that the marginal returns to agricultural production are less than the marginal costs of land production, which results in the transfer not maximizing net income, indicating uncertainty about the income gains of the transferring farmers. Compared to the farm households whose land has been transferred out, those whose land has been transferred in are more constrained by natural conditions. This is because farm households whose land has been transferred out can choose to work in the city and earn more non-farm income. In contrast, the longer benefit cycle of agriculture makes it difficult to compensate for the non-farm income gained from working in the city in the short term. More importantly, farm households whose land is transferred have a larger area to cultivate, spend more time on agricultural labor, and have less energy and time to spend on entrepreneurship and other business activities. According to Eswaran et al. [42], only when the cultivated area exceeds a certain threshold will farm households choose to hire labor for farming based on rational considerations. Otherwise, it will only increase land inputs. Farm households whose land is transfer in are more susceptible to financing constraints. Still, land transfer in is also an investment behavior (households need to spend money to rent land from other households). Therefore, it is difficult for financing-constrained farmers to make other investments and reduce their risk appetite when land is transferred in. It is difficult for farm households whose land is transferred in to profit from other sources, which is not conducive to livelihood diversification.

Hypothesis 1

(H1a): Land transfer can reduce income inequality in farm households.

Hypothesis 1

(H1b): The effect of land transfer out is higher than that of land transfer in.

Entrepreneurship is an essential channel for farm households to increase their income. Previous studies have shown that entrepreneurship can alleviate rural poverty, increase farm households’ income, reduce the urban–rural income gap [43], and reduce income inequality. Land plays an essential role in farm households’ entrepreneurship, and the network of social, political, economic, and cultural relations on land resources is a fundamental element that affects farm households’ entrepreneurship. We believe that land transfer has a good effect on entrepreneurship. Land transfer can reduce the income inequality of farm households through the role channel of farm households’ entrepreneurship, mainly in the following two aspects: First, land transfer influences the investment willingness of farm households [32] and provides partial financial support for entrepreneurship. Due to the profit-seeking nature of finance, farm households are more susceptible to financing constraints and find it challenging to start a business [44]. The funds obtained from the land transfer will increase the availability of credit to farm households, and farm households will have a more optimistic view of the market, increase their risk appetite, and be more likely to engage in entrepreneurial behavior. Second, land transfer affects the allocation of labor resources in farm households [45], freeing the household’s population from the land and providing time and labor for entrepreneurship. Since workers will pursue utility maximization after carrying out land transfers, farm households gradually withdraw from agriculture and become more involved in higher-paid non-farm work. The labor force of farm households allows families to choose employment in a broader range, which improves the efficiency of household labor resource allocation and is conducive to entrepreneurship by farm households [46]. Based on the above analysis, research hypothesis 2 is proposed.

Hypothesis 2

(H2): Land transfer will reduce the income inequality of farm households through entrepreneurship as a channel of action.

The existing literature states that various characteristics of farm households affect their income and may also affect the probability of their entrepreneurial behavior [47,48]. Farm households need access to information through the Internet when they are engaged in land transfer. If farm households obtain more information about agricultural inputs, their investment strategies may change and effectively promote land transfer out and in [49]. Access to information can also change the allocation of production and cropping structure factors, thus increasing agricultural productivity [50]. Farm households who do not use the Internet are disadvantaged in their ability to access market information and have less competitiveness in land transfer. Compared to urban areas, individuals in rural areas will be more familiar with each other, and transactions occur in rural areas so that social capital may be very important. Farm households will be more inclined to invest in modern agricultural factors of production if they have a high social capital index [51]. The informal system formed by social capital may affect farm households’ land resource allocation decisions. More robust social capital means more vital trust between the two parties, which can help farm households effectively reduce the transaction costs of land transfer, which is conducive to the conclusion of the land transfer transaction and promotes the long-term participation of farm households in the land transfer [52], and the decline in transaction costs makes it more likely that farm households will be able to obtain benefits in the process of land transfer. Social capital can also increase the performance rate of land transfer contracts because individuals with more robust social networks have higher losses due to reputation decline in the event of non-performance. Farm households are intrinsically incentivized to fulfill contractual agreements to keep their reputations from declining. Trust between the two parties to the transaction will reduce moral hazard and lower the probability of opportunistic behavior [53]. Therefore, we believe that farm households who use the Internet and have more substantial social capital may gain more from the role of land transfer in terms of income inequality. Based on the above analysis, research hypothesis 3 is proposed.

Hypothesis 3

(H3): Farm households who use the Internet and have more substantial social capital will gain more in the role of land transfer in terms of income inequality.

3. Data and Methods

3.1. Data Sources

The data used for analysis in this paper are matched data. There are two primary sources. The first is 2018–2020 data from the China Family Panel Study (CFPS), a long-term, fixed-sample tracking survey hosted by Peking University’s China Center for Social Science Surveys. The CFPS focuses on the economic and non-economic well-being of the Chinese population, as well as on several research topics, including economic activity, educational attainment, family relationships and household dynamics, population migration, and health. The CFPS is a nationwide, large-scale, multidisciplinary social tracking survey program. The explanatory variable we focus on is income inequality among farm households, so we deleted the urban sample and variables with missing values. Finally, 27,134 samples entered our empirical analysis. The second source is Chinese provincial macro data from the National Bureau of Statistics of the People’s Republic of China. We extracted the urbanization rate, GDP per capita, and industrial structure data for the provinces included in the CFPS2018-2020 sample for the corresponding years. The sample distribution is shown in Figure 1.

3.2. Methods

3.2.1. Selection of Model Variables

The variable explained in this study is income inequality. There are many ways of measuring income inequality. Common ones are the Thiel index and the Gini coefficient. However, traditional income inequality measures such as the Theil index and the Gini coefficient can only reflect income inequality within a group. They cannot reflect relative income inequality at the individual level, making it difficult to identify causal relationships at the micro-level. In contrast, the Kakwani Index can reflect the state of relative income deprivation at the micro-individual level and thus make up for this shortcoming [54]. In this study, we choose the Kakwani index to measure income inequality. The Kakwani index has been used to measure income inequality for many years, and many scholars have applied it for calculations [55,56], so we think this measure is reasonable. The Kakwani index is calculated as follows:

$$\begin{array}{c}RD\left(y_i\right)={\displaystyle \frac{1}{n{\mu }_Y}}\left[{\displaystyle \sum _{j=i+1}^n} \left(y_j-y_i\right)\right]={\gamma }_{y_i}^{+}\left({\displaystyle \frac{{\mu }_{y_i}^{+}-y_i}{{\mu }_Y}}\right)\end{array}$$

(1)

where μ_Y denotes the mean value of the income of all sample farm households in cluster Y, μ+ y_i denotes the mean value of the income of the sample farm households in the cluster whose income exceeds y_i, and γ+ y_i denotes the percentage of the sample farm households in the cluster whose income exceeds y_i in the total sample farm households. The Kakwani index is calculated in the range of [0,1], and the larger the index value, the higher the degree of inequality in the income of farm households.

The explanatory variable for this study is land transfer. There should be two forms of land transfer: out and in. Referring to some previous literature, if farm households make either one of land transfer out and in, we regard it as the existence of land transfer [57,58]. This generates a binary variable that is assigned a value of 1 if the farm household has carried out a land transfer and a value of 0 if it has not carried out a land transfer. In addition, we will carry out an additional discussion by regressing land transfer out and in separately.

The three mechanism variables chosen for this study are entrepreneurship, Internet use, and social capital. Referring to Djankov et al.’s study, we define entrepreneurship as self-employment or starting a private business by a member of the entire household [59]. Internet use was measured in terms of whether or not the household head used a cell phone or computer to access the Internet, referring to the study by Zhang et al. [60]. China is a typical humane society, and the social network formed through human interaction plays an important role in people’s lives and is one of the most important ways for residents to acquire social capital. Therefore, concerning Cao’s study, we measure social capital as the logarithm of expenditures on favors and gifts [61]. Specifically, entrepreneurship is the mediating variable in this paper and, therefore, is regressed using the traditional three-step test for mediating effects. While Internet use and social capital are the moderating variables, the interaction term between them and land transfer is brought into the regression. The moderating effect holds if the interaction term is significant [62].

Several control variables are also selected in this paper to minimize the error in the regression. There are three primary levels of control variables. The first is the essential characteristics of farm households (including age, gender, education, and a series of other variables) [63]. We also include the squared term for age. This aims to explore whether age has a U-shaped relationship with income inequality. This is a common practice in econometrics [64]. The second is household characteristics (including family size and working outside the home) [65]. The third is provincial characteristics (including urbanization rate, GDP per capita, and industrial structure) [66]. The industrial structure is the ratio of tertiary sector output to secondary sector output. It is generally believed that the more developed the tertiary industry in a region, the higher the level of industrial structure [67]. Descriptive statistics are shown in

Table 1. Descriptive statistics.

Variable	Definition and Measure	Mean	Std. Dev.
Income inequality	Kakwani index	0.438	0.226
Land transfer	Whether the farm household is engaged in land transfer or not. Yes = 1, otherwise = 0.	0.248	0.432
Land transfer out	Farm household transfer out of land Yes = 1, otherwise = 0.	0.154	0.361
Land transfer in	Farm household transfer in of land Yes = 1, otherwise = 0.	0.102	0.303
Gender	Gender of head of household. Male = 0, Female = 1	0.498	0.500
Age	Age of head of household	41.086	14.179
Age squared	Age of head of household squared	1889.12	1223.973
Edu	Educational level of the head of household. Illiterate = 1, Elementary school = 2, Junior high school = 3, Senior high school = 4, College = 5, Bachelor’s degree = 6, Master’s degree = 7, Doctorate = 8	2.800	1.273
Work status	Whether the head of the household is working. Yes = 1, otherwise = 0.	0.731	0.443
Medical insurance	Whether the farm household has medical insurance. Yes = 1, otherwise = 0.	0.907	0.291
Social security	Whether the farm household has social security. Yes = 1, otherwise = 0.	0.609	0.488
Family size	Number of persons in farm household	4.468	2.046
Labor migration	Whether the farm household has labor migration. Yes = 1, otherwise = 0.	0.511	0.499
Urban	Urbanization rate in the province where the farm household is located	57.121	9.282
Gdp	GDP per capita of the province where the farm household is located	10.840	0.376
Ins	Level of industrial structure (tertiary/secondary) in the province where the farm household is located	1.294	0.373
Entrepreneurship	In the past 12 months, did any member of your household engage in self-employment or start a private business? Yes = 1, otherwise = 0.	0.102	0.303
Internet usage	Does the head of household use a computer or cell phone to access the Internet? Yes = 1, otherwise = 0.	0.616	0.486
Social capital	Value of money and gifts in kind spent by all household members in the last 12 months (logarithms)	7.222	2.383

.

3.2.2. The Models

According to the existing literature, and in conjunction with the research purpose of this paper, OLS is constructed as follows:

$$\begin{array}{c}{Y}_{it}={\beta }_0+{\beta }_1{LF}_{it}+{\beta }_2{X}_{it}+{\epsilon }_i\end{array}$$

(2)

In Equation (1), Y represents income inequality, the variable explained in this study, i represents different farm households, and t represents time. β0 represents the intercept term. LF represents the explanatory variable of this study, land transfer, and β1 represents the coefficient of the effect of land transfer on income inequality. Xi is the full control variable, and β2 is the coefficient of the effect of the control variable on income inequality. ε is an error term.

4. Results

4.1. Baseline Regression

The results of the baseline regression are presented in

Table 2. Regression to baseline.

	Income Inequality
Variables	(1)	(2)	(3)	(4)
Land transfer	−0.026 ***	−0.030 ***	−0.026 ***	−0.029 ***
	(0.003)	(0.003)	(0.003)	(0.003)
Gender		0.009 ***	0.007 ***	0.007 ***
		(0.003)	(0.003)	(0.002)
Age		−0.007 ***	−0.006 ***	−0.005 ***
		(0.001)	(0.001)	(0.001)
Age squared		0.000 ***	0.001 ***	0.001 ***
		(0.000)	(0.000)	(0.000)
Edu		−0.042 ***	−0.050 ***	−0.043 ***
		(0.001)	(0.001)	(0.001)
Work Status		−0.023 ***	−0.030 ***	−0.027 ***
		(0.003)	(0.003)	(0.003)
Medical Insurance		−0.017 ***	−0.006	−0.019 ***
		(0.005)	(0.005)	(0.005)
Social Security		−0.021 ***	−0.020 ***	−0.032 ***
		(0.003)	(0.003)	(0.003)
Family size			−0.026 ***	−0.028 ***
			(0.001)	(0.001)
Labor migration			−0.035 ***	−0.044 ***
			(0.003)	(0.003)
Urban				0.001 ***
				(0.000)
GDP				−0.167 ***
				(0.007)
Ins				−0.010 **
				(0.004)
_cons	0.471 ***	0.729 ***	0.880 ***	2.608 ***
	(0.002)	(0.013)	(0.013)	(0.068)
Fixed effect	YES	YES	YES	YES
Observations	27,134	27,134	27,134	27,134
R2	0.030	0.118	0.183	0.232
Mean VIF	1.83

Note: ** p < 0.05, *** p < 0.01; robust standard errors in parentheses.

. Column 1 of Table 2 is the result of the regression, including only the effect of land transfer on the income inequality of farm households. Column 2 of Table 2 results from adding demographic characteristics such as gender and age. Column 3 results from adding household characteristic variables (family size and labor migration). Column 4 is the result of adding all control variables. The average VIF value is less than 10, indicating the absence of multicollinearity. We can find that the impact coefficients of land transfer are negative in the results of all four regressions and pass the significance test at the 1% level. This shows the negative impact of land transfer on income inequality among farm households. Hypothesis 1a of this study is established. Suppose we focus on the results of the control variables. In that case, we can find that education level, whether or not they work, social security, and medical insurance effectively reduce income inequality among farm households. What is interesting is the age variable; the impact coefficient of age is negative, but the square impact coefficient of age is positive, indicating that the increase in age first reduces income inequality. Still, after reaching a certain level, it increases income inequality. Regarding household characteristics, the impact coefficients of household size and labor migration are negative and pass the significance test at the 1% level, suggesting that both variables can reduce income inequality in farm households. Some characteristics of the city may also affect the income of farm households; we can find that the urbanization rate, GDP per capita, and industrial structure significantly affect the income inequality of farm households. The increase in GDP per capita and industrial structure can effectively reduce the income inequality of farm households. Still, the development of urbanization will increase the income inequality of farm households.

Land transfer can be categorized into land transfer out and land transfer in. According to previous studies, the roles of these two approaches may differ. We regress land transfer out and land transfer in separately, and the regression results are shown in

Table 3. Results of land transfer out and in.

	Income Inequality
Variables	(1)	(2)
Land transfer out	−0.043 ***
	(0.003)
Land transfer in		0.001
		(0.004)
Control variables	YES	YES
Fixed effect	YES	YES
Observations	27,134	27,134

Note: *** p < 0.01; robust standard errors in parentheses.

. We find from columns (1) and (2) of Table 3 that the coefficient of land transfer out is −0.043, and the coefficient of land transfer in is 0.001, and that land transfer out passes the test of significance at the 1% level, while land transfer in does not pass the test of significance. This indicates that land transfer out has a much higher impact on income inequality of farm households than land transfer in. Hypothesis 1b of this study is established.

4.2. Robustness Test

Robustness tests are conducted in this study to ensure the reliability of the baseline regressions. The robustness test of this study consists of several aspects: 1. The effect of land transfer on income is tested using quantile regression. If at lower income quartiles, land transfer can show a stronger effect on income improvement, while at higher income quartiles, land transfer can show a lower effect on income improvement, it can be used to prove that land transfer reduces income inequality in another way. 2. Replacement of other models. The results of the benchmark regression may change with the model’s replacement. Suppose the effect of land transfer on income inequality does not change significantly with the model’s replacement. In that case, it means that the results of the benchmark regression are robust. 3. Consider endogeneity. Endogeneity may lead to errors in the results of the benchmark regression, so it is necessary to re-run the regression based on the 2SLS model using instrumental variables.

4.2.1. Quantile Regression

The results of the quantile regressions are shown in

Table 4. Results of quantile regression.

	Income
	Q25	Q50	Q75
Variables	(1)	(2)	(3)
Land transfer	0.123 ***	0.110 ***	0.076 ***
	(0.012)	(0.012)	(0.013)
Control variables	YES	YES	YES
Fixed effect	YES	YES	YES
Observations	27,134	27,134	27,134

Note: *** p < 0.01; robust standard errors in parentheses.

. For the sake of aspect analysis, here we replace income inequality with the total income of the farm household in a year (logarithmic). We can see that the coefficients of land transfer in columns (1), (2), and (3) are 0.123, 0.110, and 0.076, respectively, which all pass the test of significance at 1%. This result suggests that land transfer raises more income for farm households in the lower-income group while having less impact on the higher-income group. Therefore, the result in Table 4 proves, on the other hand, the effect of land transfer on income inequality of farm households.

4.2.2. Propensity Score Matching

Since the explanatory variables in this study are binary variables, the benchmark model can be replaced by taking propensity score matching (PSM) and re-running the regression. The results of PSM are shown in Column 1 of

Table 5. Results of the robustness test.

	Income Inequality	Income Inequality	Land Transfer	Income Inequality
	PSM	Tobit	2sls	2sls
Variables	(1)	(2)	(3)	(4)
Land transfer	−0.033 ***	−0.029 ***		−0.081 ***
	(0.004)	(0.002)		(0.018)
IV			0.976 ***
			(0.040)
F-statistics			711.674
Fixed effect	YES	YES	YES
Observations	27,134	27,134	27,134	27,134

Note: *** p < 0.01; robust standard errors in parentheses.

. We can find that the coefficient of the effect of land transfers on income inequality among farm households under the PSM model is −0.033, which passes the test of significance at the 1% level; this indicates that the benchmark baseline results are robust.

4.2.3. Tobit Model

The explained variable in this study is an index synthesized through the Kakwani method, which takes values between 0 and 1, so it can be identified as a restricted variable. A common regression model for restricted variables is the Tobit model, so we use the Tobit model for further robustness tests. The regression results of the Tobit model are shown in Column 2 of Table 5. We can find that the coefficient of land transfer on income inequality among farm households under the Tobit model is −0.029, which passes the test of significance at the level of 1% and is comparable to the benchmark regression. The difference in results is not significant. This indicates that the results of the baseline regression are robust.

4.2.4. 2SLS Model

Although we have conducted many robustness tests before, the results of the benchmark regression still face some challenges. The results of the benchmark regression are not necessarily accurate due to the possibility of reverse causality. Therefore, it is a common practice in academia to seek an instrumental variable. We combined the previous studies and conducted a regression based on the 2SLS model using the average land transfer values of other farm households in the same region as instrumental variables. The regression results are shown in Columns 3 and 4 of Table 5. We can find that the benchmark regression results still hold after considering the endogeneity problem due to reverse causality. The coefficient of the effect of land transfer on income inequality among farm households is −0.081, which passes the test of significance at the 1% level, and the F-statistic value is 711.674, which is greater than the critical value of the weak instrumental variable.

4.3. Heterogeneity Test

Land transfer’s impact may differ for farm households in different regions. This is because natural conditions, farm households’ habits, crops grown, and levels of economic development are different in different areas. These factors may affect land transfer. So, to understand the differences in the role of land transfer in different regions, we conducted a heterogeneity test. The regression results are shown in

Table 6. Results of heterogeneity regression.

	East	Mid	West	North	South	Non-Food-Producing Areas	Food-Producing Areas
Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Land transfer	−0.030 ***	0.003	−0.040 ***	−0.039 ***	−0.015 ***	−0.024 ***	−0.031 ***
	(0.004)	(0.005)	(0.004)	(0.003)	(0.004)	(0.004)	(0.003)
Control variables	YES	YES	YES	YES	YES	YES	YES
Fixed effect	YES	YES	YES	YES	YES	YES	YES
Observations	10,956	6498	9680	16,070	11,064	13,565	13,569

Note: *** p < 0.01; robust standard errors in parentheses. The eastern region includes Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong and Hainan; the middle region includes Shanxi, Anhui, Jiangxi, Henan, Hubei and Hunan; and the western region includes Neimenggu, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu, Qinghai, Ningxia and Xinjiang. The southern region includes Sichuan, Guizhou, Yunnan, Chongqing, Hainan, Guangxi, Guangdong, Hunan, Hubei, Jiangxi, Fujian, Anhui, Zhejiang, Jiangsu and Shanghai; the northern region includes Beijing, Tianjin, Hebei, Shanxi, Neimenggu, Henan, Shandong, Liaoning, Jilin, Heilongjiang, Shaanxi, Gansu, Qinghai, Ningxia and Xinjiang.

. From columns (1), (2), and (3), the coefficients of the impact of land transfer are −0.030, 0.003, and −0.040, respectively, and the coefficient of column (2) does not pass the significance test. In contrast, the other two columns pass the significance test at the 1% level. This suggests that land transfer can effectively reduce income inequality for farm households in the east and the west, but the effect is insignificant for farm households in the center. From the results of columns (4) and (5), the impact coefficients of land transfer are −0.039 and −0.015, respectively, which pass the significance test at the 1% level. This indicates that land transfer can effectively reduce the income inequality of farm households in the north and the south, but the impact on farm households in the north is higher than in the south. From the results of columns (6) and (7), the impact coefficients of land transfer are −0.024 and −0.031, respectively, and all of them pass the significance test at a 1% level. This indicates that land transfer can effectively reduce the income inequality of farm households in non-food-producing and food-producing areas. Still, the effect is greater for farm households in food-producing areas. The possible reason for this is that the western region is in the beginning or accelerated development stage of large-scale operation. Compared to the east and the middle, the marginal effect of land transfer is naturally more pronounced. On the other hand, the east has superior natural conditions and economic development. It is more conducive to agricultural mechanization and large-scale cultivation than the central part of the country. Compared with northern China, land transfer is more active in the south; for example, the transfer rate in Fujian exceeds 30%, while the rate of land transfer in the north is lower. As a result, the marginal effect from land transfers is stronger for farmer households in the north, and the role of land transfers is greater than in the south. Compared to non-food-producing areas, farmer households in food-producing areas are more likely to operate on a large scale, and land transfer is conducive to the expansion of production, making land transfer more useful. Farmer households in non-food-producing areas are more inclined to increase their income by working outside the home rather than through land.

4.4. Mechanism Test

In the previous section, we verified the impact of land transfer on farm households and conducted robustness tests. However, as we mentioned in the theoretical analysis, we would like to know the channel of action through which land transfer causes an impact on income inequality. Therefore, we conducted a mechanism test, and the results are presented in

Table 7. Results of mechanism tests.

Variables	Entrepreneurship	Income inequality	Income Inequality	Income Inequality
	(1)	(2)	(3)	(4)
Land transfer	0.032 ***	−0.025 ***	−0.029 ***	−0.031 ***
	(0.004)	(0.002)	(0.002)	(0.002)
Entrepreneurship		−0.117 ***
		(0.004)
Land transfer × Internet usage			0.006 ***
			(0.002)
Land transfer × Social capital				0.007 ***
				(0.001)
Fixed effect	YES	YES	YES	YES
Control variables	YES	YES	YES	YES
Observations	27,134	27,134	27,134	27,134

Note: *** p < 0.01; robust standard errors in parentheses.

. Columns (1) and (2) of Table 7 show that the coefficient of land transfer on entrepreneurship is 0.032, which passes the significance test at the 1% level. The coefficient of land turnover on income inequality remains significant after adding the entrepreneurship variable to the baseline regression. This suggests that land transfer can impact farm household income inequality through the channel of action of promoting entrepreneurship. Hypothesis 2 of this study is established.

Column (3) of Table 7 shows that the coefficient of land transfer’s effect on farm households’ income inequality is −0.029, which passes the significance test at the 1% level. The coefficient of the interaction term between land transfer and Internet use is 0.006, which passes the significance test at a 1% level. This indicates that the effect of land transfer on income inequality among farm households is positively moderated by Internet use. Farm households that use the Internet gain more from land transfer. Column (4) of Table 7 shows that the coefficient of land transfer’s effect on farm households’ income inequality is −0.031, which passes the significance test at a 1% level. The coefficient of the interaction term between land transfer and social capital is 0.007, which passes the significance test at the 1% level. This indicates that the effect of land transfer on income inequality among farm households is positively moderated by social capital, and farm households with stronger social capital gain more from land transfer. Hypothesis 3 of this study is established.

5. Discussion

Using data from the China Family Panel Study, this study finds that land transfer reduces income inequality among farm households. While previous studies have focused on the fact that land transfer can increase the income of farm households, they do not understand whether it impacts income inequality. Due to the separation of the three rights over land in rural China, land transfer is neither a land purchase nor a land lease [68]. This is because ownership is neither in the hands of the buyer nor the seller. The transaction between the two parties only involves a contractual right (right of disposal) and an operation right, so land transfer is a unique phenomenon in China. Other countries are more likely to allow the buying or leasing of land [69]. Therefore, the literature on the area of land and farmers’ income inequality has focused on land use and land distribution policies [70,71]. The difference behind this may be due to the issue of land ownership. Land in China is not privately owned. However, in most countries, land, especially rural land, is privately owned. As the most direct resource for farm households, land is closely related to their lives. Research related to land suggests land will directly affect farm households’ lives and needs to be emphasized. In addition, research on income inequality is more in line with today’s goal of inclusive growth than focusing on income improvement. The experience of developed countries shows that rural areas are more likely to be left behind in economic development and face income inequality [72]. Thus, this study also enriches the research related to income inequality by providing cases from developing countries. In addition, an essential difference between this study and previous studies is that it clarifies the difference between land transfer in and land transfer out on income inequality. Previous studies have only shown that both can increase incomes but have not discussed the impact on income inequality [31,32]. Clarifying the difference is essential to help us better understand the mechanisms by which farm incomes increase during land transfers and promote sustainable growth. Interestingly, this study also found that the effect of land transfer on income inequality among farm households is only effective for farmers in the east and west and has a negligible impact on farmers in the center. This study also has some limitations and we may have missed some important variables. For example, variables such as agricultural subsidies, government intervention, and digital finance may affect the inequality of farm household income. Based on this issue, the coefficients of the benchmark regression may change.

6. Conclusions and Policy Implications

Based on data from the 2018–2020 China Family Panel Study (CFPS), we empirically investigate the impact of land transfer on income inequality of farm households by building an OLS model and draw the following conclusions: First, land transfer significantly reduces income inequality of farm households. This conclusion remains robust after using different estimation methods and considering endogeneity issues. Additionally, the role of land transfers out is greater than the role of land transfers in. Second, the impact of land transfer on farm household income inequality varies by region: in terms of east, center, and west, farm households in the east and west gain more from land transfer than those in the center. From the south and north perspectives, farm households gain more from land transfer than in the south. From the perspective of food-producing areas, farm households gain higher returns than those in non-food-producing areas. Finally, entrepreneurship, using the Internet, and social capital are important mechanisms by which land transfer affects income inequality among farm households. The effect of land transfer on farm household income inequality is transmitted through the mediating variable of entrepreneurship. However, whether or not farm households use the Internet and whether or not they have stronger social capital have an impact on the effect of land transfers on income inequality. Farm households using the Internet and with stronger social capital will gain more from land transfer.

Based on the above conclusions, the following policy recommendations are put forward: (1) The government should continue to promote land transfer in rural areas, treat land transfer as a fundamental project, improve the market system for the transfer of rural land property rights, and enhance the efficiency of rural land factor allocation. A social insurance system related to various land transfer modes, such as agricultural insurance and subsidies, should be established and improved to fully protect the interests of farm households. At the same time, agricultural enterprises are encouraged to develop moderate-scale operations through land transfer, enhance agriculture’s competitiveness and development quality, and assist farm households in increasing their income and realizing common prosperity. (2) Adequate research and adherence to local conditions: Rural situations vary greatly, and guidance and policy support should be provided by region and category. The government should give full consideration to the land resources and diversification conditions of each village and combine them with the economic development trend of each region to guide the centralized development of rural land, transfer it in an orderly and reasonable manner, and promote the movement of capital, technology, and other factors to the countryside. For villages with sufficient natural resource conditions, services should be optimized, basic conditions should be set, and public opinion should be fully respected. However, for regions unsuitable for land transfer, the transfer should not be forced through administrative orders because of the pursuit of political achievements. (3) Make full use of digital technology for publicity in land transfer. The government can popularize relevant policy information through videos on the Internet, TV, and cell phones. Through multi-faceted, deep-level publicity, the villagers’ awareness of each land transfer model is increased, thereby increasing the support for the land model.