Impact of Urbanization on Cropping Structure: Empirical Evidence from China

Abstract

Food security is a national priority and a cornerstone for maintaining national and regional stability. Focusing on cropping structure directly related to food security, this paper estimates the causal effect of urbanization on cropping restructuring in China. We use provincial panel data between 2000 and 2019 and threshold regression models to identify potential structural mutation characteristics. This study is an attempt to fill the cognitive gap for the nonlinear relationship between urbanization and cropping restructure. Urbanization formed agricultural labor supply constraints that significantly increased the share of sown area of grain crops, but with obvious threshold characteristics, and the effect of urbanization on cropping structure was no longer significant after crossing the threshold. Urbanization in the main grain-producing regions and main grain-selling regions promoted the adjustment of cropping structure in the direction of grain but was not significant in balanced production and marketing regions. Short-distance urbanization did not bring significant effects on cropping restructuring. We propose several suggestions for cultivated land planting structure, urbanization, and food security.

1. Introduction

Promoting a new type of people-centered urbanization is an important part of China’s modernization path [1,2]. With the deepening of the reform of the urban–rural dual household registration system and the improvement of the social security system, the equalization of basic public services in urban and rural areas has been preliminarily realized, various obstacles for the transfer of rural labor force to cities have been broken, and urbanization has developed rapidly. Public data shows that the urbanization rate increased from 17.9 percent in 1978 to 65.2 percent in 2022 and has already entered middle and late stage. Urbanization has promoted the transfer of the rural labor force from the agricultural sector with low productivity to the non-agricultural sector with high productivity, and its social identity, livelihood as well as welfare have all undergone significant changes and improvements, which has provided an important driving force for China’s economic development [3,4,5]. However, urbanization has accelerated the feminization, child-oriented, and aging structure in rural China, which has gradually tightened the constraints on the supply of the agricultural labor force [6,7]. According to data released by the National Bureau of Statistics China, the population in rural areas dropped from 808 million in 2000 to 501 million in 2020, with an average annual decline of more than 15.35 million. During the same period, the number of people working in agriculture continued to decline from 360 million to 180 million, an average annual decline of about 9 million. In addition, according to the seventh National Census, the proportion of the elderly aged 60 and above in rural areas in 2020 will be 23.81 percent and 17.72 percent, respectively, 7.99 percentage points and 6.61 percentage points higher than that in urban regions. The population in rural areas presents obvious characteristics of decreasing number and aging structure on the whole, and the supply of an effective agricultural labor force appears to be at a weak point.

FAO data indicates that the global hungry population in 2021 is between 702 million and 828 million. The proportion of severely food insecure in the global population has increased significantly from 10.9% in 2020 to 11.7% in 2021, and many are caught in the plight of severe food insecurity, so the task of ensuring food security is arduous and significant. Agricultural cropping structure is an important factor to determine food security [8,9], and there are obvious differences in planting and agronomy among different crop types. Among them, grain crops and economic crops have obvious differences in labor demand, with the former being significantly less than the latter. According to data released by the National Bureau of Statistics China, the average labor consumption of rice, wheat, corn, and other grain crops was 4.44 days per mu in 2020, while that of open-air economic crops was 30.94 days per mu, about 6.97 times that of grain crops. The labor consumption of facility economic crops was 55.47 days/mu, about 12.49 times that of grain crops. Meanwhile, grain crops are easier to mechanize and have obvious advantages in labor substitution. From the actual situation, the relative “standardization” of grain crop planting agronomy is more suitable for mechanical operations, while the mechanization of economic crops is relatively difficult. Statistics show that in 2019, the overall mechanization rate of rice, corn, and wheat exceeded 80 percent, with rice at 81 percent, corn at 88 percent, and wheat at 95 percent. Comparatively speaking, economic crops are limited by the problem of operation standardization, and the overall mechanization rate is at a low degree. For example, 28% of cotton crops were harvested by machinery, while only 10% were harvested in the main sugarcane-producing regions in 2017. Since grain crops and economic crops show different characteristics of labor demand, under the condition of agricultural labor supply constraints formed by urbanization, a cropping structure adjustment has become a rational choice to alleviate the constraints of agricultural labor supply.

Then, to what extent will the constraints of agricultural labor supply caused by urbanization affect the adjustment of agricultural cropping structure? What are the characteristics and heterogeneity of this effect? Further, how can China formulate targeted policies and measures to ensure food security in the process of urbanization? To systematically answer the questions above, this paper conducts an empirical analysis of the above questions based on large sample provincial panel data.

The rest of this paper is organized as follows: Section 2 provides a literature review; next, Section 3 introduces model selection, variable setting, and data source; Section 4 discusses the empirical results; Section 5 gives a further analysis for related issues; and, finally, the last section outlines the conclusions and policy implications.

2. Literature Review

Rapid urbanization has produced higher GDP and brought an increase in rural household income and the diversification of income sources [10,11,12,13,14]. However, a large number of agricultural labor force off-farm transfer has gradually formed constraints on the supply of the agricultural labor force, especially the outflow of the young and middle-aged labor has caused serious problems of aging, feminization, and child-oriented structure in rural regions [15,16]. Under the influence of urbanization, production factors such as capital, labor force, and farmland in household agricultural operations are inevitably reallocated [17,18], which has an impact on economic decisions, livelihood strategies as well as agricultural planting structure [19,20]. According to the existing research, there are mainly the following two logical routes. Firstly, due to labor supply constraints brought by urbanization, farmers tend to participate in the cultivation of non-labor-intensive grain crops and reduce their participation in the production of relatively labor-intensive economic crops, or adopt extensive management strategies, such as reducing the multiple-cropping index and abandoning crops and wasteland in remote plots, resulting in the loss of economic advantages in farmland use [21,22,23,24]. Secondly, the increase in wage income of rural households brought about by urbanization can effectively alleviate the liquidity constraints faced by farmers’ households [25] and help improve farmland investments as well as abilities to resist risks [26,27], thus encouraging farmers to shift their production focus to economic crops with a higher return on investment. (According to the data of the National Agricultural Cost and Benefit Information Compilation 2019, the average net profit of the three major staple grains (rice, wheat, and corn) in 2018 was CNY −85.59/mu, while the average net profit of economic crops (vegetables, for example) reached CNY 2690.93/mu.) According to the logic above, the relationship between urbanization and agricultural cropping structure adjustment depends on the strength of the influence of labor supply and capital inflow on decision-making behavior. When the constraints of agricultural labor supply become tight and wage income is not heavily invested in agricultural production, urbanization may promote the adjustment of agricultural cropping structure to “grain-oriented”. However, when the income of migrant workers is enough to alleviate the impact of labor supply constraints on the adjustment of agricultural planting structure, the impact of urbanization on the cropping structure to grain-oriented may be limited.

Numerous studies have carried out rich discussions and research on the relationship between urbanization and agricultural cropping structure adjustment, but no consensus conclusion has been reached. Liu et al. [28] believed that rural labor’s participation in non-agricultural employment significantly increased income and eased household liquidity constraints, thus helping to increase the proportion of economic crops in all crops. Liu et al. [29] found that urbanization has significantly changed cropping structure, with most of these regions showing rapid increases in the proportion of fruit and vegetables. Huang et al. [30] found that the increase in non-farm employment opportunities brought about by urbanization contributes to the increase in the income level of rural residents. As a result, food consumption shifts and is upgraded from a focus on sustenance to a focus on food safety and nutrition, which is manifested as a decrease in the direct market consumption of food products and a strong demand for high value-added agricultural products such as animal protein in the form of pork, poultry, beef, and dairy products [31,32,33]. Since the rising cost of planting grain crops reduces the comparative returns of grain production and economic crops have higher comparative returns, rational farmers would adjust their farming behavior and planting structure in response to market demand and price changes, thus increasing the possibility of planting economic crops. In addition, some studies have emphasized that the objective function of farmers has changed in the process of urbanization, from the cultivation of “food safety first” grain crops in the past to the cultivation of economic crops with higher net income, thus leading to the “non-grain-oriented” agricultural planting structure [34].

Contrary to the above research conclusion, some scholars believe that urbanization increases the proportion of grain crop sown area. Specifically, because grain crops are characterized by a high degree of mechanization with less labor input and easy participation in the agricultural division of labor, the constraints on agricultural labor supply caused by urbanization will prompt farmers to plant grain crops [35,36]. Especially in the case of the gradual rise and continuous improvement of agricultural machinery socialization services in rural areas [37], the positive influence of urbanization on the adjustment of agricultural cropping structure to grain-oriented will be gradually strengthened [38]. Li et al. [39] found that urbanization have led to the weakening of the agricultural labor force, while forcing farmers to accept new agricultural technologies to save labor inputs and promote the development of agricultural mechanization, and thus, to facilitate the restructuring of planting towards food crops. Wang et al. [40] further studied and found that labor out-migration for work would prompt peasant households to invest their income in machinery to alleviate labor supply constraints, so the planting probability and planting area proportion of grain crops would not be changed. However, when the out-migrating labor force is the head of the household, the proportion of grain crops will be significantly increased in rural households, thus correspondingly reducing the proportion of economic crop rate. Another part of the literature found that the improvement of agricultural socialization service and mechanization can effectively replace the agricultural labor force and ease labor supply constraints so that the feminization and aging of the rural labor force will not make a significant impact on the agricultural planting structure adjustment [41].

The existing literature provides a rich theoretical basis and experience reference for this research, but there are still issues to be further explored and clarified, which are embodied in the following three aspects. Firstly, since the adjustment of agricultural cropping structure is a rational decision-making behavior of farmers under the constraints of labor supply, when urbanization is in a state of change and adjustment, labor migration and the corresponding supply constraints are bound to have dynamic characteristics. Therefore, in different development stages of urbanization, whether its effect on the adjustment of cropping structure can always maintain a linear relationship needs to be further verified. Secondly, in theory, there may be a reverse causality between urbanization and cultivated land cropping structure adjustment, that is, potential endogenous problems are faced in the investigation of the effect of urbanization on cropping structure, and the empirical results will be biased. Therefore, it is necessary to adopt targeted measurement methods to alleviate the endogeneity and try to obtain more reliable research conclusions. Thirdly, there is a vast territory in China, as well as great differences in resource endowment, urbanization pattern, and agricultural cropping structure characteristics among different regions, which may lead to inconsistency between the empirical results of regional and overall samples. Therefore, it is necessary to systematically investigate heterogeneity to provide theoretical support for the formulation and implementation of differentiated policies.

Given this, this research tries to make a new attempt from the following three aspects. First, the threshold regression model was used to estimate the structural mutation points of urbanization affecting the adjustment of agricultural cropping structure, and to grasp the dynamic change characteristics of the relationship between them. Secondly, to find the effective instrumental variables of urbanization, and to obtain reliable parameter estimates through the use of the two-stage least square (2SLS) method. Thirdly, according to the classification criteria of grain functional areas, the total samples were divided into main grain-producing regions, main grain-selling regions, and balanced production and marketing regions, and the potential heterogeneity of urbanization affecting cropping structure was further investigated in an attempt to expand the existing research conclusions to a certain extent and provide a theoretical basis for the differentiation policies.

3. Empirical Strategies

3.1. Methods

To investigate whether the impact of urbanization on the agricultural cropping structure adjustment shows characteristics of structural mutation, this paper refers Hansen [42], and sets a threshold regression model as follows:

$$Y_{it}={\beta }_0+{\beta }_{1}Urba{n}_{it}+{\beta }_{2}X+{\tau }_{it}, \mathrm{if} q_i\le \gamma$$

(1)

$$Y_{it}={\phi }_0+{\phi }_{1}Urba{n}_{it}+{\phi }_{2}X+{\eta }_{it} \mathrm{if} q_i>\gamma$$

(2)

where $Y_{it}$ represents the agricultural cropping structure; $Urba{n}_{it}$ is the key independent variable of interest, the urbanization at provincial level; $X$ is a series of control variables that may affect the cropping structure of interest; ${\beta }_0$, ${\beta }_1$, ${\beta }_2$ and ${\phi }_0$, ${\phi }_1$, ${\phi }_2$ are the parameters to be estimated or the parameter vectors; $q_i$ is the threshold variable, $\gamma$ is the threshold value to be estimated; ${\tau }_{it}$ and ${\eta }_{it}$ are error terms and obey independent homo-distribution.

In fact, the model is equivalent to a piecewise function: when $q_i\le \gamma$, the coefficient of the key independent variable is ${\beta }_1$; when $q_i>\gamma$, the coefficient of urbanization is ${\phi }_1$, and the coefficient of other control variables also changes simultaneously.

Considering that the proportion of the sown area of grain crops is a limited variable ranging from 0 to 1, negative fitting values may be generated if the traditional linear method is used to estimate the model directly, and the estimated values of parameters may be biased. Therefore, the panel data tobit model following several studies [43,44] dealing with limited dependent variable will be used for a robust check in this paper. The specific form of function is as follows:

$$Y_{it}={\alpha }_0+{\alpha }_{1}Urba{n}_{it}+{\displaystyle \sum _j{\beta }_j{x}_{j,it}+{\mu }_i+{\eta }_{it}}$$

(3)

where $Y_{it}$ is the proportion of grain crop sown area at the provincial level; $Urba{n}_{it}$ is the urbanization rate of i province in t year; $x_{j,it}$ represents control variables; ${\mu }_i$ is the individual effect; ${\eta }_{it}$ is the random disturbance term; and ${\alpha }_0$, ${\alpha }_1$, and ${\beta }_j$ are the parameters to be estimated.

3.2. Variables

The dependent variable is cropping structure. This paper mainly investigates whether the agricultural cropping structure has been adjusted to grain-oriented under the constraints of agricultural labor supply. Therefore, the dependent variable was expressed as the proportion of the sown area of grains (including rice, wheat, and corn) in total. It should be noted that according to the statistical standard of the National Bureau of Statistics China, food crops include not only rice, wheat, and corn, but also soybeans and tubers. However, considering that the agronomy of soybean and potato is quite different from that of cereal, the corresponding logic of adjustment may not be consistent with that of cereal. In addition, the state has put forward a food security strategy of “basic self-sufficiency of grain and absolute security of staple food”, with an emphasis on grains and rations. Therefore, it is more consistent with the overall logic of this paper to measure the adjustment of agricultural cropping structure by the proportion of the grain-sown area to the total, which also matches the basic national food security strategy.

The independent variable is urbanization. In the context of the long-term existence of the urban–rural dual structure, the rural labor force continues to flow to the city and realize urbanization, especially some young and middle-aged labor force shows stronger mobility, thus forming a rigid constraint on the supply of agricultural labor force, which is also the main mechanism that theoretically causes the adjustment of agricultural cropping structure. This index was expressed as “the proportion of the permanent urban population in total”. In addition, to investigate the influence of geographical location, distance, and other characteristics of labor transfer on the adjustment of agricultural planting structure, “the proportion of the sum of the number of employments in rural private enterprises and the number of individual employees in rural population” is used to measure the index of short-distance urbanization for relevant expansion analysis.

Following Li et al. [45], Huang et al. [30], and considering the availability of data, control variables were introduced from three dimensions: population characteristics, production conditions, and external environment. Among them, the population characteristics include sex ratio, dependency ratio, and education level. The production conditions include three variables: the quantity of agricultural machinery, irrigation condition, and farmers’ income. The external environment includes two variables: road network density and industrial structure. These variables are selected according to the following.

Firstly, farmers’ decision-making behavior of grain planting is based on the behavioral motivation of pursuing maximum income [46]; meanwhile, the agricultural labor force has the characteristics of non-homogeneity, that is, the human capital contained in different labor forces and the labor intensity they can bear are different. Therefore, it is necessary to control some demographic characteristics at the provincial scale. In the empirical model, three variables including sex ratio, dependency ratio, and education level are mainly introduced at the provincial level. Specifically, (1) sex ratio reflects the gender structure within the province. A higher sex ratio means a more male population and a more effective labor supply, which may further affect the agricultural planting structure. (2) Dependency ratio reflects the age structure in the province, and the high dependency ratio means that the age structure in the province is biased towards childhood and aging. On the one hand, it can directly affect the supply of an effective agricultural labor force. At the same time, the care needs of the dependent groups have a crowding effect on the family labor force, which will indirectly affect the supply of the agricultural labor force from the perspective of time allocation and may further affect the agricultural planting structure. On the other hand, when the family dependency is relatively high, it may cause the spatial locking of the family labor force, that is, the labor force cannot leave the household, which will increase the supply of effective labor force, and thus, bring a negative impact on the adjustment of the agricultural planting structure to the grain-oriented structure. (3) Education level reflects the quality of the labor force, is an important factor to determine farmers’ individual behavior as well as decision-making [47], and may affect the adjustment of the agricultural planting structure.

Secondly, agricultural machinery, irrigation condition, and farmers’ income variables were selected. Extant studies have pointed out that agricultural mechanization can effectively reduce agricultural production costs, and save agricultural labor input [48,49]. From the perspective of the variables related to agricultural production conditions, the mechanized operation level of grain crops is significantly higher than that of economic crops [50], and the higher amount of agricultural machinery may help to adjust the agricultural planting structure toward food crops. Agriculture is a water-intensive industry [51]; namely, water resources are an important element in agriculture. Irrigation conditions reflect agricultural irrigation conditions, and irrigation water affects agricultural planting structure directly [52]. Irrigation water requirements in several types of crops are significantly different [53,54]. For economic crops, especially some vegetable crops, water consumption is relatively high, which requires higher irrigation conditions. Higher irrigation conditions may help to increase the planting proportion of economic crops, thus promoting the agricultural planting structure toward non-grain-oriented. However, irrigation conditions play an important role in the location movement of planting structures. For example, major grain-producing areas in China are gradually concentrated in perennial and supplementary irrigated areas in the north, which may further increase the proportion of food crops sown. Income increase helps to relax the liquidity constraints faced by farmers, which is helpful to improve agricultural investment, and push cropping structure towards non-grain-oriented. On the other hand, under the realistic situation that wage income accounts for more and more of farmers’ income, the opportunity cost of agricultural production and operation keeps increasing, which further promotes the adjustment to a labor-saving planting structure, which may contribute to a grain-oriented structure.

Finally, external environment variables such as road network density and industrial structure were investigated. Road network density reflects transportation conditions, and the transport infrastructure can be defined as a factor that guarantees the growth and economic development of the region [55]. Compared with corn, wheat, and other grain crops, economic crops require higher traffic conditions and market location, so the proportion of economic crops planted in areas with better road infrastructure may be higher, that is, road network density may promote the adjustment of agricultural cropping structure in the direction of a non-grain-oriented structure. The industrial structure is measured by the proportion of the sum value of secondary and tertiary industries in the gross regional product, which has a direct impact on the employment structure of farmers. Generally speaking, the secondary and tertiary industries have a high employment occupancy rate [56], which can provide sufficient placement space for agricultural labor to obtain non-agricultural employment [57]. The transfer of agricultural labor to non-agricultural industries will help to promote the adjustment of labor-saving planting structures, thus affecting the adjustment of agricultural cropping structures. The descriptive statistics of variables are shown in

Table 1. Descriptive statistics of the Samples.

Variable	Definition	Mean	SD	Min	Max
Cropping structure	Proportion of grain crops area to all crops	0.5348	0.1270	0.0345	0.8568
Urbanization	Proportion of urban resident population in total	0.5018	0.1563	0.2190	0.8960
Sex ratio	Ratio of the number of males to the number of females	1.0448	0.0383	0.9225	1.2317
Dependency ratio	Sum of child support ratio and old-age support ratio	0.3691	0.0740	0.1732	0.5758
Education	Number of people educated at elementary school and above as a percentage of the total population	0.9113	0.0725	0.4141	0.9849
Agricultural machinery	Power of agricultural machinery per unit of sown area (kW/mu)	0.3849	0.2221	0.0878	1.6417
Irrigation condition	Proportion of effective irrigated area in cultivated land area	0.5089	0.2217	0.1404	0.9902
Farmers’ income	Per capita disposable income of rural residents (Chinese yuan), based on CPI excluding price change factors	73.1940	53.9286	13.2733	323.8556
Road network density	Ratio of road mileage to provincial area (kilometers/square kilometers)	0.7139	0.4923	0.0188	2.1159
Industrial Structure	Proportion of secondary and tertiary sector values in regional GDP	0.8782	0.0661	0.6209	0.9973

.

3.3. Data Sources

The data used in this paper include 620 benchmark samples of 31 provinces in China from 2000 to 2019, mainly from various statistical yearbooks. Specifically, data on the structure of agricultural cultivation were obtained from the China Statistical Yearbook (2001–2020); data on urbanization from 2000 to 2004 were calculated from the China Demographic Yearbook (2001–2005), and data from 2005 to 2019 were obtained from the China Statistical Yearbook (2006–2020). Among the other control variables, the road network density variable is highlighted, as this variable involves the arable land area data, the base data are obtained from the China Statistical Yearbook, CSMAR, and the Information Network of the Development Research Center of the State Council. In addition, other control variables are obtained from the China Statistical Yearbook and the China Rural Statistical Yearbook.

4. Empirical Results

4.1. Data Check

In this article, we check the heteroskedasticity with a modified Wald test, which corresponds to a p-value of 0.000, indicating the existence of heteroskedasticity. The Wooldridge test and Pesaran’s test were used to check autocorrelation in time series as well as in spatial, and their corresponding p-values are 0.000 and 0.117, respectively. The results indicate the existence of time series autocorrelation, but cross-sectional autocorrelation was not supported. To overcome the problems of heteroskedasticity and within-group autocorrelation, this paper will use robust standard errors for clustering at the provincial level and reports the corresponding t-values. In addition, we added a multicollinearity check, and the VIF of each variable was less than 5 (VIF < 10 in general), indicating that there is no serious multicollinearity problem with the selected variables. We also check normality based on the Shapiro–Wilk test. The estimation results show that the p-value corresponding to each variable is greater than 0.05, and the null hypothesis that the data are distributed normally cannot be rejected, indicating that the data distribution follows normality. Meanwhile, by plotting the scatter plots of the dependent variables and the urbanization which this paper is interested in, we find that they have a strong linear relationship, with only some divergence in distribution as the independent variables increase, but the linear relationship is satisfied overall.

4.2. Benchmark Results

The results of the F-statistic obtained after setting a single threshold and the p-value obtained by bootstrap sampling 300 times are shown in

Table 2. Threshold regression estimation results.

Threshold Effect Test
Threshold Category	F-Statistic	p-Value (Bootstrap Sampling 300 Times)
Single Threshold Test	44.92 *	0.0967
Threshold Estimation Results
Threshold Variable	Threshold Value
	Estimated Value	Confidence Interval (95%)
Urbanization	0.8155	[0.7957, 0.8201]
Parameter Estimation Results
Independent Variables	Coefficients	t-value
Urbanization (<0.8155)	0.2832 ***	3.29
Urbanization (>0.8155)	0.1362	1.60
Sex ratio	0.0638	0.46
Dependency ratio	0.0383	0.43
Education	−0.2811 **	−2.09
Agricultural machinery	0.0442 *	1.78
Irrigation condition	0.1336 ***	3.15
Road network density	0.0500	1.40
Farmers’ income	0.0002	0.92
Industrial Structure	−0.5046	−1.22
Constant	0.9767 **	2.64

Note: Here are t-values corresponding to robust standard errors for clustering at the provincial level; ***, **, and * indicate two-tailed t-tests statistically significant at the 1%, 5%, and 10% levels, respectively.

. From the results of the threshold test, it can be found that the single threshold effect of urbanization passed the significance test at the 10% statistical level, indicating the existence of a threshold value. At the same time, the results of the test of the estimated value of the threshold show that the threshold value of urbanization is 81.55% and the value is significant at the 95% confidence level, indicating that there is a significant single threshold effect in the economic process of urbanization affecting agricultural crop restructuring, that is, the effect of the former on the latter is characterized by structural mutation.

From the estimation results of the parameters of the threshold regression model, the effect of urbanization on the adjustment of agricultural cropping structure is not consistent in different intervals; among which, when the urbanization is less than 0.8155, the effect of urbanization on the cropping structure is significant and positive. Similar studies have also shown that labor transfer will significantly increase the proportion of households engaged in grain crop cultivation [58,59] using different types of models such as the GTWR method. However, when urbanization crosses the threshold of 0.8155, the effect is no longer significant. In terms of sample distribution, Beijing, Tianjin, and Shanghai crossed the threshold of urbanization in 2005, 2012, and 2000, respectively, while other provinces have not yet crossed the threshold. Further, according to the basic logic of this research, the main mechanism by which urbanization leads to the grain-oriented farming structure is the urban–rural allocation of labor and the formation of an effective labor supply constraint in agriculture, which induces farmers to adopt labor-saving farming schemes, i.e., grain-oriented. From this, we can assume that the impact effect arises mainly from the labor force with mobility characteristics, considering that the urbanization variable used in this paper is measured by the proportion of the urban resident population to the total population, which includes a certain proportion of the less mobile household urbanized population, for example, the Hukou urbanization rate in China was 44.38% in 2019. Based on this ratio, we can further calculate the urbanization of the mobile population at the threshold to be about 37.17%. In other words, at the national level, when the urbanization rate of the rural migration population is less than 37.17%, the rural labor outflow will significantly contribute to the grain-oriented sector of the agricultural cultivation structure. However, when the urbanization level of the mobile population crosses the threshold of 37.17%, the rural labor outflow and the consequent tightening of labor supply do not have a significant impact on the restructuring of crops.

In terms of marginal effects, it is easy to find that for every 1 percentage point increase in urbanization, the share of the grain crops sown area in total increases by about 0.28 percentage points accordingly. Further, we use the result to conduct a partial equilibrium analysis where urbanization increases from 36.22% in 2000 to 60.60% in 2019, an average annual increase of about 1.28 percentage points. Then, given other factors, urbanization will contribute to an average annual increase of about 0.36 percentage points in the proportion of area sown to grain crops. Further, based on the average value of the total crop sown area of 2.390 billion mu from 2000 to 2019, the sown area of grain crops would increase by about 8.604 million mu per year. Thus, the urbanization of the population has a significant effect on the adjustment of agricultural cultivation structure to grain-oriented, and at least from the perspective of “basic self-sufficiency of grain and absolute security of staple food”, urbanization of the population helps to achieve the desired goal of the food security strategy.

In terms of control variables, the significant and negative effect of education level on the grain-oriented adjustment of agricultural cropping structure indicates that education level reduces the proportion of area sown to grain crops. This may be due to the idea that a higher education level matches the demand for higher-skilled labor for high-value economic crops, thus reducing the proportion of acreage sown to food crops. The coefficient of the agricultural machinery ownership variable is significant and positive, indicating that a higher level of agricultural mechanization can increase the proportion of acreage sown to grain crops, which may be closely related to a higher mechanization rate of grain crops when agricultural machinery ownership is higher; farmers may choose to grow a higher proportion of food crops to maximize the dilution of fixed costs and reduce redundancy of machinery inputs. The coefficient of the irrigation condition variable is significant at the 1% statistical level, indicating that effective irrigation conditions can significantly contribute to a higher proportion of sown area for grain crops. This may be because areas with better irrigation conditions have correspondingly higher irrigation efficiency, which pushes the main grain-producing areas to concentrate in the northern perennially irrigated and supplementally irrigated areas, which in turn pushes up the proportion of sown area of grain crops.

4.3. Robustness Check

To verify the robustness of the results of the benchmark model, this paper excludes the sample with an urbanization rate greater than 81.55% and performs parameter estimation based on the remaining sample using the panel tobit model and fixed effects model. It should be noted that there are two commonly used models regarding panel data tobit regression: the mixed tobit regression model and the random effects tobit regression model. In general, a benchmark determination principle in choosing a specific model is based on the likelihood ratio test results, and

Table 3. Robustness check results.

Variables	Tobit Model		Fixed Effect Model
	Coefficient	z-Value	Coefficient	t-Value
Urbanization	0.0861 *	1.81	0.0876 *	1.80
Sex ratio	−0.1109	−1.30	−0.0867	−1.00
Dependency ratio	0.0109	0.23	0.0229	0.47
Education	−0.2852 ***	−3.41	−0.2847 ***	−3.25
Agricultural machinery	−0.0087	−0.32	−0.0126	−0.44
Irrigation condition	0.0301	0.67	0.0372	0.75
Road network density	0.0462 ***	3.80	0.0477 ***	3.82
Farmers’ income	0.0004 ***	3.87	0.0004 ***	3.76
Industrial structure	−0.5924 ***	−5.63	−0.6098 ***	−5.66
Constant	1.3120 ***	8.97	1.2930 ***	8.72
FE	YES		YES
Within R2	—		0.1731
Hausman test p	—		0.0011
ρ	0.8580 *** (0.0329)		—
LR χ2	800.06		—
Log-likelihood	857.21		—
Observations	577		577

Note: *** and * denote two-tailed t-tests statistically significant at the 1% and 10% levels, respectively; the observation here is 577 because samples with urbanization levels greater than 81.55% were excluded.

reports the estimation results of each parameter of the model and the correlation test. From the likelihood ratio test results, the individual effect is significant, and the ρ-value (ρ is variance ratio coefficient, representing the proportion of individual effect variance in the total variance of compound error) is greater than 0.8, indicating that the individual effect plays an important role in explaining the agricultural cropping structure adjustment, and the individual effect needs to be considered in the model. Table 3 reports the estimation results of the tobit as well as fixed effects model for panel data.

The results in Table 3 show that the log-likelihood value of the tobit model is 857.21, and the corresponding p-value is 0.0000, indicating that the model fits well overall. From the specific coefficient estimates, the estimated coefficient of urbanization is significant and positive at the 10% statistical level, which indicates that urbanization significantly increases the proportion of sown area for grain crops, i.e., it promotes the adjustment of agricultural cultivation structure to grain-oriented. From the estimation results of the fixed effects model, it can be seen that urbanization has a significant positive effect on the share of sown area of grain crops, which is consistent with the results of the baseline regression.

4.4. Examining Potential Endogeneity Issues

In examining the effect of urbanization on the adjustment of agricultural cropping structure to grain-oriented, we may face the problem of endogeneity, which leads to biased and inconsistent parameter estimation results. Specifically, the possible cause of the endogeneity problem is reverse causality: With a relatively stable amount of agricultural land, farmers can optimize the allocation of household labor through an agricultural cropping structure adjustment. Specifically, regions with a higher proportion of food crop cultivation area have a greater substitution effect on agricultural labor due to a higher level and efficiency of mechanized operations, which may thus help to contribute to the outflow of farm household labor and the urbanization of the population. To overcome the possible endogeneity problem, this paper attempts to find the instrumental variables of the key independent variable and use two-stage least squares (IV-2SLS) regression to estimate the effect of urbanization on the grain-oriented farming structure. Specifically, this paper chooses lagged one-period urbanization as an instrumental variable [60]. Since one-period lagged urbanization is an outcome that has already occurred, the current period agricultural cropping structure adjustment does not affect it, thus blocking the two-way causality problem. Meanwhile, the time series of urbanization has some autocorrelation, thus satisfying the correlation condition of the instrumental variable.

Table 4. Estimation results of instrumental variables.

Variables	Stage I		Stage II
	Coefficient	t-Value	Coefficient	z-Value
Urbanization	—	—	0.1593 **	2.38
Lagged urbanization	0.7257 ***	28.94	—	—
Sex ratio	−0.0969 **	−2.21	−0.0733	−0.85
Dependency ratio	−0.0751 ***	−2.86	0.0500	0.96
Education	0.0814	1.51	−0.2391 **	−2.27
Agricultural machinery	−0.0189	−1.27	0.0034	0.12
Irrigation condition	−0.0202	−0.80	0.0395	0.81
Road network density	0.0161 **	2.42	0.0458 ***	3.44
Farmers’ income	0.0003 ***	5.60	0.0003 **	2.49
Industrial structure	0.1006 *	1.74	−0.7112 ***	−6.25
Constant	0.0930	1.09	1.2814 ***	0.1658
FE	—		YES
Within R2	0.9115		0.1741
Weak IV test F-value	581.40		—
Observations	547		547

Note: ***, **, and * denote two-tailed t-tests statistically significant at the 1%, 5%, and 10% levels, respectively; samples with an urbanization rate greater than 81.55% are excluded.

reports the estimation results of each parameter of the model.

From the results of the instrumental variables test, the F-value of the first stage weak instrumental variables test is significantly greater than the usual criterion (F-value = 10), indicating that there is no weak instrumental variables problem. Considering the endogeneity issue, the coefficient of urbanization is significant and positive in the direction of a grain-oriented structure, i.e., urbanization helps to increase the proportion of the sown area for food crops, and the magnitude and significance of the coefficient (from 10% to 5%) are significantly higher compared to the previous fixed effects model, but the direction of the effect remains the same. In addition, the results further validate the robustness of the baseline model results and the reliability of the study findings.

5. Further Expansion Analysis

5.1. Grain Functional Regions Inspection

Although the above results can reveal the overall impact of urbanization on the agricultural cropping structure adjustment, they ignore the typical fact that cropping structures and resource endowments vary greatly among different regions in China. In order to leverage regional comparative advantages and ensure food security, China’s 31 provinces were divided into main grain-producing regions with 13 provinces, main grain-selling regions with 7 provinces, and balanced production and marketing regions with 11 provinces since the 1990s. The main grain-producing regions are suitable for growing food crops, in terms of geography, soil, and climate, as well as other natural conditions, with high grain yield and a large proportion of cultivation to ensure self-sufficiency while also being able to transfer grain out as a commodity. In 2022, the main grain-producing regions including Heilongjiang, Jilin, Liaoning, Inner Mongolia, Hebei, Henan, Shandong, Jiangsu, Anhui, Jiangxi, Hubei, Hunan, and Sichuan accounted for more than 78% of the total grain production, bearing a major responsibility to ensure national food security. The main grain-selling regions are relatively economically developed, but with more people and less land, there is a large gap between grain production and demand, including in Beijing, Tianjin, Shanghai, Zhejiang, Fujian, Guangdong, and Hainan. The balanced production and marketing regions make limited contributions to the national grain production but can basically maintain self-sufficiency, including in Shanxi, Ningxia, Qinghai, Gansu, Tibet, Yunnan, Guizhou, Chongqing, Guangxi, Shaanxi, and Xinjiang.

In general, there are systematic differences in different stages of urbanization, agricultural cultivation structure, and grain production factor structure among grain functional regions. To characterize the heterogeneity of urbanization influencing the agricultural planting structure adjustment in each grain functional area, this study will investigate the main grain-producing regions, main grain-selling regions, as well as balanced production and marketing regions in sub-samples. It should be noted that to ensure unbiased and consistent estimation results, we keep the samples with urbanization less than the threshold and estimate the model using two-stage least squares, and the estimation results are shown in

Table 5. Results of grain-functional regions samples.

Variables	Main Grain-Producing Regions		Main Grain-Selling Regions		Balanced Production and Marketing Regions
	Stage I	Stage II	Stage I	Stage II	Stage I	Stage II
Urbanization	—	0.0812 *	—	0.8820 ***	—	−0.1287
		(1.89)		(6.70)		(−1.43)
Lagged urbanization	0.8862 ***	—	0.7511 ***	—	0.6376 ***	—
	(25.27)		(9.52)		(14.46)
Sex ratio	−0.0402	−0.1492	−0.0399	0.0688	−0.0731	−0.0466
	(−1.06)	(−1.13)	(−0.39)	(0.54)	(−0.70)	(−0.34)
Dependency ratio	−0.0324	0.1320 *	−0.2026 **	0.2474 **	−0.1002 *	−0.2229 ***
	(−1.56)	(1.81)	(−2.19)	(1.99)	(−1.78)	(−2.96)
Education	0.0217	−0.1408	0.2120	−0.2400	0.0997	−0.0694
	(0.46)	(−0.86)	(0.98)	(−0.87)	(1.03)	(−0.54)
Agricultural machinery	0.0101	0.0377	0.0887	0.2971 **	−0.0306	−0.0097
	(0.69)	(0.75)	(0.94)	(2.49)	(−1.05)	(−0.25)
Irrigation condition	0.0196	0.1102 *	−0.0775	0.0064	−0.0378	0.0725
	(1.13)	(1.83)	(−1.20)	(0.08)	(−0.55)	(0.81)
Road network density	0.0011	0.0498 ***	−0.0364	0.1057 **	0.0262	0.0129
	(0.20)	(2.63)	(−1.06)	(2.41)	(1.64)	(0.60)
Farmers’ income	0.0001 **	0.0009 ***	0.0002	0.0006 ***	0.0007 ***	0.0004 **
	(2.06)	(4.96)	(1.45)	(3.10)	(4.72)	(2.07)
Industrial structure	0.0739 **	−0.5216 ***	−0.0798	−0.9413 ***	−0.0855	−1.1488 ***
	(2.10)	(−4.12)	(−0.31)	(−2.99)	(−0.46)	(−4.77)
Constant	0.0538	0.9913 ***	0.1659	0.9292 **	0.2407	1.6847 ***
	(0.89)	(4.68)	(0.51)	(2.31)	(1.14)	(6.04)
FE	—	YES	—	YES	—	YES
Within R2	0.9811	0.5542	0.9004	0.3446	0.8522	0.2897
Weak IV test F-value	1295.06	—	76.37	—	121.11	—
Observations	247		91		209

Note: ***, **, and * denote two-tailed t-tests statistically significant at 1%, 5%, and 10% levels, respectively; samples with an urbanization rate greater than 81.55% were excluded; t-values in parentheses in stage I, and z-values in parentheses in stage II.

.

From the results, it can be seen that the urbanization in the main grain-producing regions and the main grain-selling regions show a significant positive effect on the adjustment of the agricultural planting structure shift to grain, but the magnitude of the coefficient differs significantly, with the former being about one-tenth of the latter. This may be due to the high proportion of the sown area of grain crops in the main production areas to the total sown area of crops, which lacks the necessary adjustment flexibility in the cropping structure even when facing labor supply constraints. The proportion of grain crops sown in the main grain-selling regions is relatively low in degree, and the adjustment elasticity of the agricultural planting structure grain-oriented is greater when faced with labor supply constraints. Taking the 2019 data as an example, the proportion of the sown area of food crops in the main production area is 76% of the total sown area of crops, and the proportion of the sown area of cereals (including rice, wheat, and corn only, beans and potatoes excluded) is 66%, which is 26 percentage points and 23 percentage points higher than that of the main grain-selling regions. In addition, the coefficient of urbanization in the balanced production and marketing area is not significant, in other words urbanization does not lead to a grain-oriented adjustment of agricultural cultivation structure. This may be because the topography of the grain production and marketing balanced areas is mostly mountainous or hilly, and the level of agricultural mechanization is low, which does not achieve effective substitution of agricultural labor, and when facing labor supply constraints, the agricultural planting structure adjustment lacks the necessary agricultural technology support.

5.2. Distance Effect Examination

Spatially different distances of urbanization imply different costs of farming, and the allocation strategies of labor factors will change accordingly, thus showing differences in the impact on agricultural cultivation structure. Specifically, the measure of urbanization in the benchmark model is the proportion of the urban resident population to the total population; that is, the spatial scope of rural labor mobility is mainly focused on the long-distance movement from villages to towns (counties and cities). Meanwhile, with the in-depth promotion of the rural revitalization strategy, private enterprises, rural recreation bases, and other industries have gradually developed with the central village as the carrier, thus generating short-distance transfer of agricultural labor within or between villages and realizing non-farm employment, as well as enjoying infrastructure and public services similar to those in cities. However, it should be noted that spatially short-distance urbanization has the characteristic of “leaving the land but not the village”, and given the lower transportation costs, farm households may make full use of labor resources and farm households faceless labor supply constraints [61]. Therefore, in this context, does urbanization still lead to a change in the structure of agricultural cultivation? To this end, this paper draws on the method of Pang et al. [62], which uses the sum of rural private employment and individual employment as a share of the rural population to measure short-range urbanization and to examine its effect on the agricultural cropping structure adjustment, and

Table 6. Results of urbanization distance effect.

Variables	Tobit Model		Fixed Effect Model
	Coefficient	z-Value	Coefficient	t-Value
Short-distance urbanization	−0.0292	−0.44	−0.0314	−0.52
Sex ratio	−0.1208	−0.83	−0.0962	−0.65
Dependency ratio	0.0021	0.02	0.0106	0.11
Education	−0.2605 **	−2.01	−0.2594 *	−1.95
Agricultural machinery	−0.0177	−0.26	−0.0218	−0.42
Irrigation condition	0.0252	0.24	0.0300	0.25
Road network density	0.0542	1.46	0.0560	1.45
Farmers’ income	0.0005 **	2.16	0.0005 **	2.12
Industrial structure	−0.5613	−1.32	−0.5808	−1.47
Constant	1.3146 ***	3.15	1.2968 ***	3.41
FE	YES		YES
Within R2	—		0.1696
Hausman test p value	—		0.0073
ρ	0.8577 (0.0387)		—
LR χ2	808.86		—
Log-likelihood	855.98		—
Observations	577		577

Note: ***, **, and * denote two-tailed t-tests statistically significant at 1%, 5%, and 10% levels, respectively; the observation here is 577 because samples with an urbanization rate greater than 81.55% were excluded.

reports the estimated results of each parameter.

From the estimation results, it can be seen that the estimated coefficient of short-distance urbanization is negative in both the tobit as well as fixed effect model, and the coefficient is highly insignificant in terms of the corresponding z-value and t-value, indicating that short-distance urbanization does not bring significant impact on agricultural cropping restructuring. Although this result is not significant, it can at least indicate that agricultural cropping restructuring is closely related to labor supply status, and spatially short-distance urbanization can provide necessary labor support for agriculture through part-time employment with low transportation cost and high mobility elasticity of agricultural labor, which does not tighten agricultural labor supply constraint, and agricultural cropping restructuring will not change significantly.

6. Conclusions and Implications

This research analyzes the impact of urbanization on the agricultural cropping structure adjustment based on panel data from 31 provinces in China from 2000 to 2019 using a threshold regression model as well as a panel tobit model. It is found that there is a significant threshold effect of urbanization on the adjustment of agricultural planting structure, with a threshold value of 81.55% in the period under examination, and that urbanization significantly increases the proportion of sown area of grain crops in the interval less than the threshold value, but the share of sown area of grain crops does not change significantly with the rise of urbanization after crossing the threshold. The above effects are significantly heterogeneous, in which urbanization significantly increases the proportion of the sown area of grain crops in the main grain-producing regions and main grain-selling regions but not in the balanced production and marketing regions. Spatially short-distance urbanization does not significantly contribute to the adjustment of agricultural cropping structure toward grain, which to a certain extent confirms that the agricultural cropping structure adjustment is an adaptive behavior of farm households under labor supply constraints.

Based on the findings above, this paper proposes four policy implications. First, we should objectively understand the phenomenon of convergence of agricultural cropping structure to grain. This implies that there is no conflict of objectives between urbanization and food security, but rather strategic consistency. Second, we identify the heterogeneity of population urbanization affecting the agricultural cropping structure adjustment. There are significant differences in the impact of urbanization on the agricultural planting structure adjustment in each functional food area, and differentiated control measures for planting structure adjustment should be implemented. Third, the impact of spatial short-distance urbanization on the agricultural planting structure adjustment is emphasized. The establishment of regional labor markets and a perfect labor market network extends to villages to guide the rural labor force to transfer to nearby areas and realize local urbanization, thus forming a village-wide labor pool and providing a basic guarantee for improving the elasticity of the planting structure adjustment. Fourth, we encourage the government to support the county as an important carrier of urbanization, from urban construction as the center of gravity to rural construction as the center, to achieve local employment, to improve the flexibility and effectiveness of factor inputs, and to meet the employment process of farmers to settle and return to their hometown employment and entrepreneurship needs.

7. Discussion

Food security is a matter of national and regional social stability. This article examines the issue of cropping restructuring, which is directly related to food security, focusing on an empirical analysis of the impact of migration in rural regions on cropping restructuring based on provincial panel data in China. The main innovations of this paper are as follows: Firstly, it confirms the potential threshold characteristics of urbanization affecting the cropping structure adjustment, namely under the labor supply constraint, in which adjusting planting structure is only a phase behavior. Secondly, it analyzes the potential heterogeneity of urbanization affecting the cropping structure adjustment in main grain-producing regions, main grain-selling regions, and balanced production and marketing regions, which provides a theoretical basis for formulating and implementing differentiated policies. Thirdly, it explores the effect of labor migration distance in rural areas on the cropping structure adjustment, which can provide theoretical support for differentiated urbanization strategies.

However, there are still some shortcomings in this paper for future research to continue to improve. This study explores the stage characteristics of cropping restructuring under the labor supply constraint but does not explore the new production strategy and its impact on food security after the labor supply constraint is further tightened after urbanization crosses the threshold. Future research can continue to explore the dynamic changes in production behavior according to the logic of labor supply constraints and collect relevant data for further empirical analysis.