Rural-Urban Migration and its Effect on Land Transfer in Rural China

Abstract

Labor force rural-urban migration will lead to changes to the land use patterns of farmers. Using the survey data on dynamic migration of the Chinese labor force in 2014, iv-probit and iv-tobit models were used to analyze the impact of labor migration on the land transfer of farmers. The results show that: (1) Off-farm employment would significantly impact land transfer of farmers and the results are robust. With every 10% increase in the proportion of off-farm employment of farmers, the average probability of rent-in land of farmers decreases by 1.55%, and the average transfer in land area of farmers decreased by 1.04%. Similarly, with every 10% increase in the proportion of off-farm employment of farmers, the average probability of rent-out land of farmers increases by 4.77%, and the average transfer out land area of farmers increases by 3.98%. (2) Part-time employment also has a significant impact on land transfer of farmers, but the impact of part-time employment on land transfer in is not robust. Specifically, with every 10% increase in part-farm employment, the average probability of rent-out land of farmers increases by 7.64%, and the average transfer out land area of farmers increases by 6.85%.

1. Introduction

With the development of urbanization and industrialization, the phenomenon of land abandonment and uninherited land has appeared all over the world [1,2,3,4,5]. This could further threaten food security and hamper the achievement of the sustainable development goals. This may be especially true in countries with more people and less land [6,7]. Since ancient times, China has been faced with the problem of the contradiction between “man and land”. That is, how a limited amount of arable land can feed a growing population [8,9,10,11]. Specifically, the average size of rural households’ farmland in China is only 0.5 hectare, which is equivalent to 1/400 of that in the United States, and China has 7% of the world land to feed 22% of the world population [10]. The conflicts between people and land as well as the limitations of the rural settlement terrain in the vast hilly areas have hindered the reasonable land transfer in some rural settlements, and the fragmented and scattered land has further restricted the sustainable livelihoods of farmers and revitalization of the countryside [12,13,14,15,16]. China’s economy has been growing rapidly for more than 40 years. However, the price of rapid economic growth is a massive transfer of rural resources (including talents, funds and land incomes) to the cities [17,18,19,20,21]. Correspondingly, there are “hollow village” phenomena [22,23,24] and abandoned land in some remote rural settlements [25,26,27,28]. According to the study of Xu et al. [19], 12% of the arable land in 27 provinces of China has been abandoned, with an average land loss of 0.33 mu (1 mu ≈ 0.067 ha). Under the background that a large amount of the rural labor force has migrated to the cities, how to realize the reasonable transfer of arable land and give play to the moderate scale effect of land have become a difficult problem that needs to be solved urgently at present.

The impact of rural-urban migration on land transfer of farmers has always been a hotspot in the fields of geography, economics, etc. [29,30,31,32,33]. Land transfer involves the transfer direction (whether there is rent-in or rent-out land) and the transfer scale (the rent-in or rent-out land area). Some studies from the academic circles have focused on the correlations between rural-urban migration and land transfer, but the empirical studies of the correlations between rural-urban migration and the direction and the scale of land transfer are fewer [34]. Meanwhile, in the few studies, labor migration (off-farm employment) is defined as laborers’ migrant working for six months or more and staying outside of their home villages by scholars, so as to explore the correlations between rural-urban migration and land transfer, and it is generally concluded that rural-urban migration would significantly hinder rent-in land of farmers and remarkably promote rent-out land of farmers (e.g., [34,35]). For example, Ji et al. [34] have found that with every 1% increase in the ratio of off-farm employment, the probability of rent-out land of farmers will increase by 8.4%, while the probability of rent-in land of farmers will decrease by 4.9%; Huang et al. [35] have discovered that when there is one more migrant worker, the probability of rent-out land of farmers will increase by 4%, whereas off-farm employment has no significant impact on rent-in land of farmers.

The above studies can increase our understanding of the impact of labor migration on land transfer. Nevertheless, the conclusion that farmers’ off-farm employment is conductive to promoting land transfer cannot well interpretation why the proportion of off-farm employment is getting higher and higher whereas the land has not been concentrated quickly enough for large-scale farmers, and the possible reason is that the part-time employment plays a crucial role [21]. Part-time employment refers to the labor force engaged in agricultural production and non-agricultural work in a year [36,37]. However, as can be seen from the existing literatures, the impacts of part-time employment on land transfer are mostly still remaining at the theoretical level, and there are few quantitative empirical studies [38,39]. Meanwhile, even in existing quantitative studies, the effect of part-time employment on land transfer is also contradictory [40,41,42].

Some studies have believed that part-time employment would accelerate land transfer (e.g., [42,43]). For example, Li et al. [44] have theoretically analyzed and argued that part-time work would change the investment structure of farmers and therefore accelerating land transfer; Xu et al. [45] has also showed that if there are more laborers working in the county and living at home for some time (similar to part-time employment), the probability of their rent-out land will be greater. Some studies have assumed that part-time employment would inhibit land transfer (e.g., [46,47]). For example, Huang [47] has considered that part-time employment of farmers would restrict land transfer and constrain the development of land scale operation market. Some studies have suggested that part-time employment may not necessarily result in land transfer (e.g., [41,42]). For example, Zhang and Qian [42] have indicated that part-time employment may not necessarily result in land transfer, and whether the land is transferred after part-time employment of farmers depends on the operating size of the land, the differences between non-agricultural income and agricultural income and so on.

In general, the existing studies can help us better understand the relationship between rural-urban migration and land transfer of farmers. However, due to the existence of part-time laborers, the special group, it is inappropriate to simply assume that labor migration would accelerate/inhibit the land transfer, and it is urgent to further explore the impact of rural-urban migration (especially part-time employment) on land transfer of farmers. Meanwhile, the conclusions on account of small-scale survey data may be biased as well, so it is urgent to use a larger and more representative data to reveal the effects of rural-urban migration on land transfer of farmers. New economics of labor migration believes that farmers would reasonably allocate labor resources under the goal of the household profit maximization, so as to decide whether laborers choose to migrate from the household level [48]. Under the guidance of this theoretical frame, using China as a case study, a representative data of 8031 farmers of 27 provinces (cities) in China were used to probed the effects of rural-urban migration on land transfer of farmers, so as to reveal the impact of rural-urban migration on land transfer of farmers and provide beneficial enlightenment for the formulation of policies related to the system and mechanism for making the elements of “talents, lands and funds” flow to the rural areas in “Rural Revitalization Strategy”. Under the premise of maximization of household income, the following basic hypotheses are put forward:

H1. 

If the ratios of part-time and off-farm employment are greater, the probability of rent-in land of farmers will be smaller, and the rent-in scale land of farmers will be lesser.

H2. 

If the ratios of part-time and off-farm employment are greater, the probability of rent-out land of farmers will be greater, and the rent-out scale land of farmers will be larger.

2. Data and Method

2.1. Data Source

The data used in this study are the China Labor-force Dynamics Survey (CLDS) conducted by Sun Yat-sen University in China in 2014, which mainly investigated the labor force migration status, land use, family production and other contents. Multi-stage multi-level probability sampling method (PPS sampling) was used to ensure the representativeness of the sample. This study focuses on the impact of labor migration between urban and rural areas on the land transfer of farmers, so only rural data are used. After cleaning, data from 8031 farmers in 209 villages in 27 provinces (cities) in 2014 were finally obtained for subsequent analysis.

2.2. Method

2.2.1. Definition and Data Description of the Model Variable

(1) Dependent variables

The situation of land transfer of farmers (including land transfer direction and land transfer scale) are the dependent variables concerned in this study. Therefore, the direction of land transfer is measured by two variables: whether farmers have land transferred in and whether farmers have land transferred out. Similarly, the scale of land transfer of farmers is measured by two variables: land transfer in scale of farmers and land transfer out scale of farmers. Among them, land transfer in means that farmers rent land from other farmers in the village through contract or oral contract; land transfer out refers to the farmers’ giving their land to other farmers for farming by contract or oral contract.

(2) Independent variables

Off-farm employment and part-time employment were used to measure the migration of labor force from rural to urban areas. Wherein, off-farm employment are defined in accordance with the research from Ji et al. [34]; Huang et al. [35], Che [49], Deng et al. [50,51], Liu et al. [52], Liu et al. [53], Ma et al. [54], referring to the individual laborers working outside their hometown villages for at least half a year; part-time laborers refers to the labor force engaged in agricultural production and non-agricultural work in a year. Additionally, in order to weaken the influence of missing variables on the estimation of model results, referring to the research from Ji et al. [34]; Huang et al. [35], Che [49], Shi et al. [55], this study selects head education, head age, household labors, land area, agricultural assets and fixed assets as the control variables of this study (

Table 1. The definition and descriptive statistical of the variables.

Variables	Variable Specific Definition	Mean	SDc
Dependent variables
Rent In	Whether farmers have rent-in land (0=no;1=yes)	0.09	0.29
Rent Out	Whether farmers have rent-out land (0=no;1=yes)	0.69	0.46
Area rented in	Rent-in land area of farmers (mu a)	1.25	14.20
Area rented out	Rent-out land area of farmers (mu a)	4.00	7.70
Part-time employment	The ratio of part-time labor force to household total labor force (%)	0.11	0.25
Off-farm employment	The ratio of off-farm labor force to household total labor force (%)	0.40	0.39
Head age	Household head’s age (year)	53.80	13.20
Head education	Whether household head has received a high school diploma or above (0 = no;1 = yes)	0.12	0.32
Household labors	Total household labor force (number)	2.74	1.60
Land area	Per capita contract land size of the household (mu a/person)	1.67	1.97
Agricultural assets	Per capita of current market value of all the agricultural assets that a household possesses (Wan Yuan b/person)	0.08	0.53
Fixed assets	Per capita of current market value of all the fixed assets that a household possesses (Wan Yuan b/person)	4.32	16.75

^a 1 mu ≈ 0.067 ha; ^b 1 US dollar ≈ 6.12 Yuan in year 2014; ^c SD = Standard deviation.

).

2.2.2. Econometric Model

The objective of this research is to explore the impact of rural-urban migration on the land transfer of farmers. Land transfer is mainly measured from two aspects, which are the transfer direction and the transfer scale. Wherein, whether farmers have rent-in land and whether farmers have rent-out land are two dichotomous variables, and this study intends to use the binary probit model to probe the impact of rural-urban migration on land transfer direction. Similarly, rent-in land area of farmers and rent-out land area of farmers are two continuous variables with positive skew distribution and a great deal of 0, and this study intends to adopt the tobit regression model to probe the impact of rural-urban migration on land transfer scale. The formula of econometric model constructed is shown below:

$${\mathrm{R}}_{ip}={\beta }_0+{\beta }_{1}of{f}_{ip}+{\beta }_{2}par{t}_{ip}+X^{\prime }\phi +{\delta }_p+{\epsilon }_{ip}$$

(1)

$${\mathrm{A}}_{ip}={\beta }_0^{*}+{\beta }_1^{*}of{f}_{ip}+{\beta }_2^{*}par{t}_{ip}+X^{\prime }{\phi }^{*}+{\delta }_p^{*}+{\epsilon }_{ip}^{*}$$

(2)

Among them, ${\mathrm{R}}_{ip}$ and ${\mathrm{A}}_{ip}$ represent the land transfer direction and land transfer scale of farmer i in province p, respectively; $of{f}_{ip}$ and $par{t}_{ip}$ represent off-farm employment and part-time employment, respectively; $X^{\prime }$ is a vector that contains the control variables of this study which may impact land transfer of farmers; ${\delta }_p$ and ${\delta }_p^{*}$ were used to control province characteristics; ${\epsilon }_{ip}$ and ${\epsilon }_{ip}^{*}$ represents the residual of the model; ${\beta }_0$,${\beta }_1$,${\beta }_2$,$\phi$, ${\beta }_1^{*}$,${\beta }_2^{*}$, ${\phi }^{*}$ represents the parameters to be estimated by the models, respectively.

Furthermore, considering that rural-urban migration may promote land transfer of farmers, and whereas land transfer of farmers may also impact the allocation of household labor resources (promote rural-urban migration). Therefore, off-farm employment and part-time employment may be endogenous variables. Referring to the research of Huang et al. [38], Xu et al. [39], Deng et al. [50] appropriate instrumental variables were used to model estimation. The selection of instrumental variables is based on the mean value of off-farm employment and part-time employment of other farmers in the same village except for those interviewed. That is IVoff_ip = (off₁+ off₂+…+ off_n-1)/(n−1) and IVpartip = (part1+part 2+…+part n−1)/(n−1). Finally, iv-probit model was used to probe the impact of rural-urban migration on land transfer direction, and iv-tobit model was used to probe the impact of rural-urban migration on land transfer scale. Meanwhile, Chi-square test statistics (Wald χ²) was used to determine whether the core independent variable is an endogenous variable. The estimated equation for the first stage is shown below:

$$of{f}_{ip}={\alpha }_0+{\alpha }_{1}IVof{f}_{ip}+{\alpha }_{2}IVpar{t}_{ip}+X^{\prime }\theta +{\omega }_p+{\mu }_{ip}$$

(3)

$$par{t}_{ip}={\alpha }_0^{*}+{\alpha }_1^{*}IVof{f}_{ip}+{\alpha }_2^{*}IVpar{t}_{ip}+X^{\prime }{\theta }^{*}+{\omega }_p^{*}+{\mu }_{ip}^{*}$$

(4)

The variables in Equations (3) and (4) are set in the same way as Equations (1) and (2), and Stata 14 (StataCorp. LLC, College Station, TX, USA) is used to estimate and calculate the entire process.

3. Results

3.1. Descriptive Statistics Analysis Result

As shown in Table 1, in terms of the dependent variables, 69% and 9% of farmers have land transfer out and land transfer in, and the average land transfer out area and land transfer in area are 4.00 mu and 1.25 mu, respectively. In terms of core independent variables, the ratio of the off-farm labor force and part-time labor force accounted for 40% and 11% of the household total labor force, respectively; In terms of control variables, the average age of household head was 56.4 years, and only 12% of the household head with high school or above. The average household has 2.74 labor forces, per capita contract land size of the household is 1.67 mu; Per capita of current market value of all the agricultural assets and fixed assets are 0.08 and 4.32 Wan Yuan/person, respectively.

3.2. Econometric Model Results

3.2.1. The Results of the Econometric Models for Land Transfer Direction of Farmers

Table 2. Probit models estimation results of land transfer direction of farmers.

Variables	Probit Models for whether Farmers Have Rent-In Land						Probit Models for whether Farmers Have Rent-Out Land
	Model 1	Model 2	Model 3	Model 4	Model 5		Model 6	Model 7	Model 8	Model 9	Model 10
Off-farm employment	−0.260 ***	−0.373 ***	−1.074 ***	−0.966 ***	−0.155 ***		0.403 ***	0.412 ***	1.613 ***	1.711 ***	0.477 ***
	(0.058)	(0.067)	(0.155)	(0.172)	(0.032)		(0.047)	(0.050)	(0.408)	(0.178)	(0.047)
Part-time employment	−0.213 **	−0.250 ***	−0.770 ***	−0.342	−0.055		0.485 ***	0.470 ***	2.730 ***	2.738 ***	0.764 ***
	(0.086)	(0.094)	(0.175)	(0.223)	(0.036)		(0.074)	(0.077)	(0.440)	(0.230)	(0.062)
Head age		0.057 ***		0.052 ***	0.008 ***			−0.055 ***		−0.036 ***	−0.010 ***
		(0.013)		(0.012)	(0.002)			(0.009)		(0.010)	(0.003)
Head age^2		−0.001 ***		−0.001 ***	−0.000 ***			0.001 ***		0.000 ***	0.000 ***
		(0.000)		(0.000)	(0.000)			(0.000)		(0.000)	(0.000)
Head education		−0.022		0.019	0.003			0.024		−0.026	−0.007
		(0.066)		(0.067)	(0.011)			(0.052)		(0.046)	(0.013)
Land area		−0.034		−0.038	−0.006			0.780 ***		0.644 ***	0.180 ***
		(0.068)		(0.114)	(0.018)			(0.129)		(0.102)	(0.028)
Household labors		0.099 ***		0.126 ***	0.020 ***			0.011		−0.071 ***	−0.020 ***
		(0.013)		(0.015)	(0.003)			(0.011)		(0.015)	(0.004)
Ln(Fixed assets)		0.025		0.043 *	0.007 *			−0.050 ***		−0.062 ***	−0.017 ***
		(0.024)		(0.024)	(0.004)			(0.018)		(0.016)	(0.005)
Ln(Agricultural assets)		0.953 ***		0.865 ***	0.139 ***			−0.890 ***		−0.374 ***	−0.104 ***
		(0.101)		(0.103)	(0.016)			(0.110)		(0.117)	(0.033)
Constant	−1.656 ***	−3.236 ***	−1.487 ***	−3.090 ***			0.451 ***	0.343	−0.027	−0.353
	(0.209)	(0.427)	(0.191)	(0.485)			(0.126)	(0.382)	(0.188)	(0.360)
Province dummies	Yes	Yes	Yes	Yes	Yes		Yes	Yes	Yes	Yes	Yes
Instrumental variables	No	No	Yes	Yes	Yes		No	No	Yes	Yes	Yes
Wald χ2	269.378 ***	458.559 ***	337.629 ***	492.424 ***	492.424 ***		929.240 ***	1002.882 ***	2179.432 ***	2373.026 ***	2373.026 ***
Endogenous Wald χ2			32.968 ***	15.814 ***	15.814 ***				58.191 ***	90.002 ***	90.002 ***
Observations	7994	7994	7994	7994	7994		8031	8031	8031	8031	8031

Note: Robust standard errors are shown in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

shows the probit models estimation results of land transfer direction of farmers. Specifically, model 1 to model 5 shows probit models for whether farmers have rent-in land. Among them, model 1 represents the result that only adding focus variables (off-farm employment and part-time employment) and controlling the province dummy variables, model 2 represents the result that including all the variables setting in Table 1. Iv-probit models were used to estimate model 3 to model 5, and the marginal effects of variables were reported in model 5. Similarly, model 6 to model 10 shows probit models for whether farmers have rent-out land. Among them, model 6 and model 7 reports the estimation results that not adopting instrumental variable method; while model 8, model 9 and model 10 reports the results that estimated by instrumental variable method, and model 10 reports the marginal effect result of variables. According to the test statistics results of Endogenous Wald χ2 in Table 2 (p < 0.01), off-farm employment and part-time employment are endogenous variables, and iv-probit models were appropriate to be used to estimate the results.

In terms of the impact of labor migration on whether farmers have rent-in land, off-farm employment is significantly negatively correlated with whether farmers have rent-in land, and the results are robust. Specifically, with every 10% increase in off-farm employment, the probability of rent-in land decreases, on average, by 1.55% (Model 5); however, part-time employment is negatively related to whether farmers have rent-in land, but the results are not robust. Specifically, in Model 1 to Model 3, part-time employment is obviously negatively correlated with whether farmers have rent-in land, while in Model 4 and Model 5, part-time employment is negatively related to whether farmers have rent-in land, but the results are not significant. With regard to the impact of labor migration on whether farmers have rent-out land, off-farm employment and part-time employment are both significantly positively correlated with whether farmers have rent-out land, and the results are robust. Specifically, with every 10% increase in off-farm employment and part-time employment, the probability of rent-out land of farmers increases, on average, by 4.77% and 7.64%, respectively (Model 5).

Additionally, there is a significant inverse U-shaped correlation between head age and whether farmers have rent-in land, and there is a significant U-shaped correlation between head age and whether farmers have rent-out land; head education is not significantly correlated with rural households’ land transfer direction; land area is not significantly related to whether farmers have rent-in land, while it is significantly positively correlated with whether farmers have rent-out land; household labors, fixed assets and agricultural assets are significantly positively related to whether farmers have rent-in land while significantly negatively correlated with whether farmers have rent-out land.

3.2.2. The Results of the Econometric Models for Land Transfer Scale of Farmers

Table 3. Tobit models estimation results of land transfer scale of farmers.

Variables	Tobit Models for Rent-In Land Area of Farmers						Tobit Models for Rent-Out Land Area of Farmers
	Model 11	Model 12	Model 13	Model 14	Model 15		Model 16	Model 17	Model 18	Model 19	Model 20
Off-farm employment	−11.664 ***	−14.173 ***	−46.712 ***	−22.210 ***	−0.104 ***		2.961 ***	1.498 ***	8.086 ***	8.197 ***	0.398 ***
	(3.706)	(3.609)	(15.144)	(5.767)	(0.027)		(0.250)	(0.217)	(1.520)	(0.586)	(0.024)
Part-time employment	−12.676 **	−10.117 **	−29.310	−5.553	−0.026		3.794 ***	2.376 ***	14.825 ***	14.097 ***	0.685 ***
	(5.630)	(4.320)	(20.752)	(7.973)	(0.037)		(0.436)	(0.374)	(1.827)	(0.715)	(0.029)
Head age		2.181 **		1.187 ***	0.006 ***			−0.215 ***		−0.186 ***	−0.009 ***
		(0.872)		(0.420)	(0.002)			(0.031)		(0.027)	(0.001)
Head age^2		−0.024 ***		−0.014 ***	−0.000 ***			0.002 ***		0.002 ***	0.000 ***
		(0.008)		(0.004)	(0.000)			(0.000)		(0.000)	(0.000)
Head education		3.669		1.805	0.008			0.388		−0.212	−0.010
		(4.027)		(2.064)	(0.009)			(0.256)		(0.194)	(0.009)
Land area		−2.496		−1.851	−0.009			5.302 ***		4.717 ***	0.229 ***
		(2.546)		(1.678)	(0.008)			(0.372)		(0.349)	(0.017)
Household labors		4.354 ***		3.470 ***	0.016 ***			1.136 ***		0.595 ***	0.029 ***
		(1.098)		(0.544)	(0.002)			(0.062)		(0.045)	(0.002)
Ln(Fixed assets)		1.235		1.543 **	0.007 **			−0.100		−0.281 ***	−0.014 ***
		(1.100)		(0.763)	(0.004)			(0.070)		(0.071)	(0.003)
Ln(Agricultural assets)		64.083 **		32.508 ***	0.151 ***			−4.325 ***		−1.765 ***	−0.086 ***
		(26.336)		(4.270)	(0.021)			(0.910)		(0.377)	(0.018)
Constant	−86.880 ***	−146.240 ***	−83.827 ***	−84.506 ***			3.265 ***	−6.103 ***	1.852 **	−5.704 ***
	(27.166)	(47.558)	(23.212)	(15.203)			(0.617)	(1.223)	(0.723)	(1.153)
Province dummies	Yes	Yes	Yes	Yes	Yes		Yes	Yes	Yes	Yes	Yes
Instrumental variables	No	No	Yes	Yes	Yes		No	No	Yes	Yes	Yes
Wald χ2			243.942	155.786	155.786				5007.226 ***	1.3e+04 ***	1.3e+04 ***
Endogenous Wald χ2			156.493	549.125	549.125				156.493 ***	549.125 ***	549.125 ***
Observations	8031	8031	8031	8031	8031		8031	8031	8031	8031	8031

Note: Robust standard errors are shown in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

shows the tobit models estimation results of land transfer scale of farmers. The settings for all models are similar to those in Table 2. According to the test statistics results of Endogenous Wald χ2 in Table 3 (p < 0.01), off-farm employment and part-time employment are endogenous variables, and iv-tobit models were appropriate to be used to estimate the results.

In terms of the impact of labor migration on land rent-in scale of farmers, off-farm employment is significantly negatively correlated with rent-in land area of farmers, and the results are robust. Specifically, with every 10% increase in off-farm employment, the average area that rent-in land area of farmers decreases by 1.04% (model 15); however, part-time employment is negatively related to rent-in land area of farmers, but the results are not robust. Specifically, in model 11 and model 12, part-time employment is significantly negatively correlated with rent-in land area of farmers, while in model 13, model 14 and model 15, part-time employment is negatively related to rent-in land area of farmers, but the results are not significant. With regard to the impact of labor migration on land rent-out scale of farmers, off-farm employment and part-time employment are both significantly positively correlated with rent-out land area of farmers and the results are robust. Specifically, with every 10% increase in off-farm employment and part-time employment, the average area that rent-out of farmers increases 3.98% and 6.85%, respectively (model 20).

Additionally, there is a significant inverse U-shaped correlation between head age and rent-in land area of farmers, and there is a significant U-shaped correlation between head age and rent-out land area of farmers; head education is not significantly correlated with land transfer scale of farmers; land area is not significant related to land rent-in area, while it is significantly positively correlated with rent-out land area of farmers; household labors, fixed assets and agricultural assets are significantly positively related to rent-in land area of farmers while significantly negatively correlated with rent-out land area of farmers.

3.3. Robustness Check

Considering that the impact of part-time employment on land transfer of farmers are not robust. Two robustness testing strategies were used for subsequent analysis. Firstly, referring to the study of Xu et al. [10], Xie et al. [36] and Liu et al. [53], part-time income and off-farm income as a percentage of total household income are used to divide part-time employment and off-farm employment. Secondly, the four dependent variables of the models (land transfer direction and land transfer scale) are approximately regarded as the continuous variables, and ivreg method were used to estimate the model results. The final results are presented in

Table 4. Robustness check models estimation results of land transfer direction of farmers.

Variables	Robustness Check Models for whether Farmers Have Rent-in Land					Robustness Check Models for whether Farmers Have Rent-out Land
	Measure of Labor Transfer by Income (Probit)		Measure of Labor Transfer by the Number of Labor Force (IVreg)			Measure of Labor Transfer by Income (Probit)		Measure of Labor Transfer by the Number of Labor Force (IVreg)
Off-farm employment	−0.562 ***	−0.528 ***	−0.174 ***	−0.137 ***		0.860 ***	0.838 ***	0.622 ***	0.628 ***
	(0.066)	(0.069)	(0.038)	(0.037)		(0.051)	(0.052)	(0.066)	(0.067)
Part-time employment	−0.198 ***	−0.123 **	−0.138 **	−0.051		0.381 ***	0.342 ***	0.963 ***	0.915 ***
	(0.048)	(0.049)	(0.067)	(0.067)		(0.040)	(0.041)	(0.104)	(0.104)
Province dummies	No	Yes	No	Yes		No	Yes	No	Yes
Instrumental variables	No	No	Yes	Yes		No	No	Yes	Yes
Control variables	No	Yes	No	Yes		No	Yes	No	Yes
Wald χ2/F statistics	321.762 ***	463.774 ***	12.071 ***	14.651 ***		1141.130 ***	1170.545 ***	45.090 ***	45.194 ***
Observations	7994	7994	8031	8031		8031	8031	8031	8031

Note: Robust standard errors are shown in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

and

Table 5. Robustness check models estimation results of land transfer scale of farmers.

Variables	Robustness Check Models Rent-in Land Area of Farmers					Robustness Check Models for Rent-out Land Area of Farmers
	Measure of Labor Transfer by Income (Tobit)		Measure of Labor Transfer by the Number of Labor Force (IVreg)			Measure of Labor Transfer by Income (Tobit)		Measure of Labor Transfer by the Number of Labor Force (IVreg)
Off-farm employment	−29.124 ***	−23.316 ***	−0.353 ***	−0.231 ***		3.636 ***	3.411 ***	1.042 ***	1.071 ***
	(8.630)	(5.377)	(0.086)	(0.084)		(0.298)	(0.256)	(0.109)	(0.108)
Part-time employment	−11.738 ***	−6.383 **	−0.488 ***	−0.228		2.033 ***	1.615 ***	1.928 ***	1.831 ***
	(4.273)	(2.544)	(0.147)	(0.145)		(0.268)	(0.228)	(0.186)	(0.185)
Province dummies	No	Yes	No	Yes		No	Yes	No	Yes
Instrumental variables	No	No	Yes	Yes		No	No	Yes	Yes
Control variables	No	Yes	No	Yes		No	Yes	No	Yes
Wald χ2/F statistics			10.478 ***	12.535 ***		119.639 ***	140.957 ***	243.337 ***	326.784 ***
Observations	8031	8031	8031	8031		8031	8031	8031	8031

Note: Robust standard errors are shown in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

.

As shown in Table 4, it can be seen that regarding the impact of labor migration on land transfer direction, whether the replacement of the core independent variable (corresponding to the Robustness Test I) or the replacement of the estimation method (corresponding to the Robustness Test II) is adopted, the direction and significance of the correlations between off-farm employment and land transfer direction of farmers are both consistent with the previous empirical analysis results, and the difference merely lies in the correlation coefficients. Specifically, off-farm employment is significantly negatively and robustly correlated with whether farmers have rent-in land, whereas it is significantly positively and robustly related to whether farmers have rent-out land. Interestingly, the results of the correlations between part-time employment and land transfer direction of farmers are in line with those of the previous empirical analysis as well, and part-time employment is negatively related to whether farmers have rent-in land, but the results are not robust, whereas part-time employment is significantly positively correlated with whether farmers have rent-out land, and the results are robust.

To be specific, when this study uses the Robustness Test I to estimate the models, part-time employment is significantly negatively related to whether farmers have rent-in land; at the same time, in the first estimation model in the Robustness Test II (when dummy variables and control variables of the province are not controlled), part-time employment is significantly negatively correlated with whether farmers have rent-in land, while after dummy variables and control variables of the province are added, part-time employment is negatively related to whether farmers have rent-in land, but the results are not significant.

As shown in Table 5, it can be seen that regarding the impact of labor migration on land transfer scale, whether the replacement of the core independent variable or the replacement of the estimation method is adopted, the direction and significance of the correlations between off-farm employment and land transfer scale of farmers are both consistent with the previous empirical analysis results, and the difference merely lies in the correlation coefficients. Specifically, off-farm employment is significantly negatively and robustly correlated with land rent-in area, whereas it is significantly positively and robustly related to land rent-out area. Interestingly, the results of the correlations between part-time employment and land transfer scale of farmers are in line with those of the previous empirical analysis as well, and part-time employment is negatively related to land rent-in area, but the results are not robust, whereas part-time employment is significantly positively correlated with land rent-out area, and the results are robust. To be specific, when this study uses the Robustness Test I to estimate the models, part-time employment is significantly negatively related to land rent-in area; at the same time, in the first estimation model in the Robustness Test II, part-time employment is significantly negatively correlated with land rent-in area, while after dummy variables and control variables of the province are added, part-time employment is negatively related to land rent-in area, but the results are not significant.

4. Discussions

Using the survey data on dynamic migration of Chinese labor force in 2014, iv-probit and iv-tobit models were used to analyze the systematic impact of labor migration on the land transfer of farmers. This research mainly has two innovation points, which can make up for the shortcomings of existing research.

Firstly, when paying attention to the impact of labor migration on land transfer of farmers, this study not only focuses on the off-farm migration group but also the part-time group, which is a special group, finding that part-time employment does not consequentially result in rent-in land of farmers. Besides, this study not only pays attention to whether farmers have rent-in and rent-out land but also focuses on the correlations between labor migrations and the rent-in and rent-out land area of farmers. Compared with the previous researches, the considerations of this research are more comprehensive. Secondly, in addition to using the iv-tobit and the iv-probit model to deal with the possible endogenous issues of the models, this study further adopts the strategies of the replacement of the core independent variable (i.e., using the ratio of income to define part-time employment and off-farm employment) and the replacement of the estimation method (i.e., using ivreg) to perform the robustness tests on the existing research results. The results of the robustness tests further verify the validity and robustness of the results of the iv-tobit and iv-probit model. Thirdly, the samples of this study cover 27 provinces in rural China. Compared with the previous small-scale or several-province sample surveys, the results of this study may better reflect China’s actual situation, and the conclusions are more universal. Fourthly, although this study takes China as a case study, it analyzes the impact of labor migration on land transfer. However, its research design ideas (e.g., measures of labor migration, especially those of part-time employment) and research methods (e.g., iv-probit and iv-tobit models) are still applicable to other countries with large population and small land (e.g., Japan, Mongolia, Nepal, Ireland), which can provide beneficial enlightenment for moderate scale land management in these countries.

China has always been faced with the problem of the contradiction between “man and land”. Rural resources are being concentrated in large quantities in cities has resulted in remarkable regional differences [56,57]. Due to labor resources flow to the cities, there are the phenomena of “hollow village” and abandoned land in many remote hilly settlements. In the context of massive labor migration, who will cultivate the land (especially the abandoned land) and how skills and funds can achieve the reasonable bidirectional flow between rural and urban areas have become difficult issues for government decision-makers. The moderate scale management under land transfer is a powerful measure to handle rural land abandonment and promote talent backflow. However, the moderate scale management does not mean larger scale is better, and it is difficult to grasp the standard of moderation. At the same time, China will face the long-term contradictions between the moderate scale management and the small-scale rural household production, and the prerequisite for realizing the rational transfer of land is to properly deal with the relationship between the two. This study could provide beneficial enlightenment for the formulation of policies related to the system and mechanism for making the elements of “talents, lands and funds” flow to the rural areas in “Rural Revitalization Strategy”. Additionally, in addition to China, many countries in the world (including developed countries such as Europe and the United States) have reported land abandonment or cannot find a suitable successor, one of the most important reasons is population migration [58,59]. In this context, the rational transfer of land can be regarded as a feasible way to prevent land abandonment and ensure food security [1,3,5]. From this point of view, the results of this study can also provide some useful enlightenment for the rational use of land in the world (especially the suppression of land abandonment).

Meanwhile, there are some similarities and differences between this study and similar studies. For example, consistent with the results of Ji et al. [34]; Huang et al. [35], Kung [60], off-farm employment would significantly affect the direction and the scale of land transfer of farmers, and the results are robust. Specifically, off-farm employment is significantly negatively correlated with whether farmers have rent-in land as well as the rent-in land area of farmers, while it is significantly positively related to whether farmers have rent-out land as well as the rent-out land area of farmers. In other words, off-farm employment would accelerate rent-out land of farmers while hinder rent-in land of farmers. Unlike the findings of Zhang & Qian [42],Yao [43], Li et al. [44] that part-time employment would accelerate land transfer) and the results of He [46] and Huang [47] that part-time employment would restrict land transfer, it is indicated from this study that part-time employment is not robustly correlated with whether farmers have rent-in land as well as the rent-in land area of farmers, whereas it is significantly positively related to whether farmers have rent-out land as well as the rent-out land area of farmers. Therefore, the conclusions of this study tend to believe that part-time employment does not result in land transfer of farmers.

Additionally, some deficiencies in this study can be further made up in future studies. For instance, this study only uses the cross-section data to explore the impact of labor migration on land transfer of farmers, while labor migration and land transfer of farmers are a dynamic changing process, and the panel data can be used to further probe the causal relationship between them in the future research. Meanwhile, labor migration is a broad concept, containing a lot of contents (such as migration time, migration location), whereas this research probes the impacts of the composing proportion of the laborers on land transfer merely from the perspectives of migration time and migration income, and the future research can further probe the impact of migration location on land transfer of farmers.

5. Conclusions

Based on the above analysis, two conclusions can be briefly drawn from the study:

(1)

Off-farm employment would significantly affect the direction and the scale of land transfer of farmers, and the results are robust. With every 10% increase in off-farm employment, the probability of rent-in land of farmers decreases, on average, by 1.55%, and the average transfer in land area of farmers decreases by 1.04%; With every 10% increase in off-farm employment, the probability of rent-out land of farmers increases, on average, by 4.77%, and the average transfer out land area of farmers increases by 3.98%.

(2)

Part-time employment also has a significant impact on the direction and the scale of land transfer of farmers. However, the correlation between part-time employment and land transfer in is not robust. Specifically, with every 10% increase in part-farm employment, the probability of rent-out land of farmers increases, on average, by 7.64%, and the average transfer out land area of farmers increases by 6.85%.