The Differential Effects of Bidirectional Urban–Rural Mobility on Agricultural Economic Resilience: Evidence from China

Abstract

The bidirectional flow of population between urban and rural areas, not limited to rural-to-urban migration, influences the sustainable development of agricultural economic resilience in multiple ways. This study employs panel data from 31 provincial-level regions in China spanning 2017–2022 to comprehensively examine the impact of bidirectional urban–rural mobility on diverse dimensions of agricultural economic resilience, while further investigating its underlying mechanisms. Benchmark regression shows that the bidirectional urban–rural mobility exerts a suppressive effect on the agricultural economic resilience. Mechanism analyses indicate that such mobility contributes to strengthening agricultural economic resilience by catalyzing land-scale operational efficiency and amplifying labor productivity gains and that the advancement of smart agriculture technologies effectively mitigates the inhibitory impacts of bidirectional mobility on agricultural economic resilience. Furthermore, according to heterogeneity analysis, the mobility exerts a suppressive effect on the resistance (Res.) and reconstruction (Recons.) of agricultural economic resilience, while concurrently enhancing its restoration (Rest.). Meanwhile, the bidirectional mobility has significantly impeded the agricultural economic resilience of the eastern, central, and western regions, as well as the primary grain-producing areas, production and marketing balance areas, and the primary grain-selling areas. Further investigation reveals that the reverse mobility has a positive effect on the resistance but a negative effect on its restoration and reconstruction.

1. Introduction

The agricultural economic resilience of countries around the world currently presents a complex and diverse pattern. As a powerful agricultural nation, the United States has a complete agricultural production system, but like Japan, it is also facing the problem of an aging agricultural labor force, which is restricting the development of its agricultural economic resilience. The labor shortage in Russia has reached as many as 200,000, limiting its agricultural production efficiency. After Brexit, the United Kingdom is faced with restricted labor mobility, leading to a decrease in its agricultural economic resilience. All the above realities collectively point to the fact that agricultural economic resilience is affected by the flow of population between urban and rural areas. Indeed, the facts bear this out. According to the Statistical Yearbook released by the Food and Agriculture Organization of the United Nations (FAO) in 2022, the global agricultural employment population decreased from 885 million in 2017 to 866 million in 2022. Among them, China, as the most populous country, saw its agricultural employment population drop from 209 million in 2017 to 193 million in 2022. However, in recent years, the urban–rural population flow in China has no longer been limited to migration to cities, but a new phenomenon of “return migration” from cities to rural areas has emerged. Especially in regions with developed rural economies, the trend of population inflow and return migration is more pronounced [1]. This trend plays a vital role in the long-term sustainable development of agricultural economic resilience.

The effect of bidirectional urban–rural population mobility on agricultural economic resilience has not been conclusively determined by the literature to date. Bidirectional urban–rural population mobility, which encompasses both rural-to-urban and urban-to-rural migration, is a significant trend in population movement. Prior to 2019, the predominant trend was unidirectional rural-to-urban flow [2]. Following that, the livability of rural areas has somewhat encouraged the movement of people from urban to rural areas [3]. The United Nations estimates that by 2025, there will be 3.12 billion people living in rural areas in the world’s least developed nations, up from 3.06 billion in 2010 [4]. There are primarily three scholarly viewpoints on how labor quantity affects agricultural economic resilience: First, the resilience of the agricultural economy will suffer as a result of the labor shortage. According to Liu et al., the dramatic decline in agricultural labor brought on by the migration of working-age labor to cities presents significant obstacles to rural economic and social development [5]. Ren et al. suggested that the labor reduction brought on by population aging impedes the sustainable development of agriculture [6]. The second point of view is that the advancement of agriculture might benefit from the loss of agricultural labor. According to Gong Binlei et al., the transfer of agricultural labor can accelerate land-scale operations, boost the degree of agricultural mechanization, and propel the high-quality development of agriculture [7]. According to the third viewpoint, shifts in the labor force will affect the agricultural economy’s resilience in both positive and negative ways. Urban–rural integration has been greatly aided by increased population mobility, which can improve agricultural efficiency and break down barriers between urban and rural areas. However, it may also result in a decrease in agricultural income and farmer productivity because of a lack of rural labor [8,9,10]. Qian Wenrong et al. proposed that only when labor forces flow equitably between urban and rural areas can the “city-dominated” development model be thoroughly transformed, so as to achieve balanced and sufficient development between cities and rural areas, enhance the viability of agricultural economy, and strengthen the resilience of agricultural economy [11].

To foster the sustainable development of agricultural economic resilience and promote agricultural economic prosperity, this paper examines the influence of bidirectional urban–rural population flow on agricultural economic resilience through the lens of such population mobility, taking China as a case study. The following are the primary causes: first, the rapid growth of China’s rural economy in recent years has made rural areas more appealing and the bidirectional mobility between China’s urban and rural populations more evident [12]; second, the country’s enormous population has put China’s agriculture in a unique position where it must connect farmers and rural areas on one end and markets on the other [13]. This paper refers to the methodology which used by Jiang Jian et al. to study the impact of population aging on agricultural economic resilience, and employs a benchmark regression model and a mediating effect model to conduct an in-depth analysis of the influence of bidirectional urban–rural population flow on agricultural economic resilience in China [14]. The paper’s contribution is threefold:

First, existing studies have examined the effects of labor force changes on agricultural economic resilience through lenses including population aging and rural-to-urban migration, yet have predominantly overlooked the contributions of urban-to-rural migrants—a critical demographic segment—to sustainable development of agricultural economic resilience. Therefore, this paper uses the chain ratio of rural and urban populations as dual indicators to measure the bidirectional flow between urban and rural populations. Based on this, it explores the differential impacts of such bidirectional flows on agricultural economic resilience and its three dimensions, thereby expanding the research boundaries of bidirectional urban–rural population flow and agricultural economic resilience.

Second, the impact mechanism of bidirectional urban–rural population flow on agricultural economic resilience is examined in this paper’s mediating effect analysis from two perspectives: labor efficiency and land scale. In addition to examining the connections between labor, land, and the economy, it thoroughly broadens the research avenues for agricultural economic resilience.

Thirdly, from the perspective of smart agriculture, this paper integrates the rapid development of smart agriculture into the analytical framework for examining the impact of two-way urban–rural population flow on agricultural economic resilience. It investigates the effects of such population movements on agricultural economic resilience under the driving force of smart agriculture, aiming to provide empirical evidence for agricultural economic research.

2. Theoretical Analysis and Research Assumptions

2.1. The Direct Impact of Bidirectional Population Mobility Between Urban and Rural Areas on Agricultural Economic Resilience

According to the push–pull theory of D.J. Bague, the purpose of population movement is to improve living conditions. It is posited that the purpose of population mobility is to improve living conditions. Factors in the areas of origin that are unfavorable for improving living conditions constitute push factors, while factors in the destination areas that are favorable for improving living conditions constitute pull factors. Migration behavior results from the interaction of the push and pull factors [15]. Large numbers of people from rural areas will move into cities when urban economic development surpasses that of rural areas. Karl Heinrich Marx’s theory of economic development holds that labor and economic development are inextricably linked. The “cost-benefit” model states that when the advantages of moving to a city outweigh the disadvantages, people in rural areas will choose to move to a city; additionally, the Todaro M P’s theory of expected income in urban and rural areas can also represent the movement of people in rural and urban areas [16], it posits that micro-level individuals migrate to maximize their own utility, where expected income and employment probability serve as the primary factors influencing migrants’ migration decisions [17]. Many urban populations will decide to relocate to the countryside as a result of the recent introduction of preferential rural policies. In this scenario, risks will affect the agricultural economic system and the agricultural economic resilience. Three dimensions are used to analyze the effects of the bidirectional movement of people between urban and rural areas on agricultural economic resilience, as per the evaluation index system of agricultural economic resilience.

First, the resistance of the agricultural economic resilience is impacted by the bidirectional movement of people between urban and rural areas. According to Zhu Mande, resistance is the agricultural production system’s ability to manage risk and provide comprehensive security [18]. Reducing the rate of rural population decline could result in over development of land to meet survival needs, which would make it harder to improve the ecological environment and upset the ecological balance of the area. Many new agricultural facilities and machineries have been put on hold as a result of the inverse population flow between urban and rural areas, which has also resulted in uneven labor quality and inconsistent adoption of new agricultural technologies. External risks frequently have an abrupt, unpredictable, and quickly spreading impact [14], which has raised agricultural costs without improving agricultural economic resilience. Simultaneously, the rural population’s recovery has raised demand for agricultural products, which will lower agricultural product prices, lower farmers’ income levels, and impair their ability to manage risks associated with agriculture.

Second, the restoration of agricultural economic resilience, that is, the capacity to fix the damaged system after it has been damaged, is impacted by the bidirectional movement of people between urban and rural areas. Increasing the number of people living in rural areas will undoubtedly help the agricultural economy recover from its damage. Production stamina is insufficient due to a reduction in the labor supply which results in a number of issues with agricultural production, including rising labor costs and declining productivity [19]. Reducing the rate of rural population decline will help address the labor shortage issue in agriculture. The additional labor force can be promptly refilled for post-disaster recovery efforts when the agricultural economy experiences market shocks and natural disasters. On the other hand, based on cutting-edge scientific knowledge and wide perspectives, human resources moving back to rural areas from cities can hasten the recovery of agricultural production in the face of agricultural disasters.

Thirdly, the reconstruction of the agricultural economic resilience, that is, the system’s capacity to self-innovate after disruption, is impacted by the bidirectional movement of people between urban and rural areas. The population of rural areas has become more diverse due to the reversal of the urban–rural population flow. This includes not only traditional farmers but also workers returning from other areas and returnees who have established businesses. These people come from a variety of backgrounds, and their knowledge and abilities are naturally combined to greatly increase the potential for agricultural technological innovation. As a significant factor in fostering high-quality agricultural development, the capacity can enhance the regenerative ability of agricultural economic resilience [20].

The following theories are put forth in light of the above theoretical analysis:

Hypothesis H1:

The resilience of the agricultural economy is hampered by the bidirectional urban–rural mobility.

2.2. The Indirect Impact of Bidirectional Population Mobility Between Urban and Rural Areas on Agricultural Economic Resilience

2.2.1. Land Scale Effect

The land scale effect states that when the population of rural areas declines, the agricultural economy will be significantly impacted by the willingness of farmers to leave and the willingness of large-scale agricultural business entities to enter. The majority of the population still moves from rural to urban areas in both directions. Farmers will decide to rent out land in a way that offers higher marginal benefits after the rural population moves to the city to prevent resource waste from abandoned land. Second, rather than depending as much on agricultural labor, large-scale agricultural business entities typically opt to use automated equipment or agricultural machinery for agricultural production. The unit value of transferred land is decreasing as more people leave rural areas, which will result the equilibrium point of maximum income from large-scale agricultural operations continue to rise, and the land scale of large-scale land operators will continue to expand. The theory of economies of scale states that increasing the scale of agricultural land management helps to make the agricultural economy more resilient.

First, on the one hand, increasing the scale of agricultural land management facilitates the use of contemporary agricultural technology, mechanization, and automation equipment, which improves agricultural production efficiency [21]. On the other hand, it can be accomplished through large-scale production, centralized procurement, labor division, and collaboration to lower the cost of agricultural production [22]. Accordingly, by lowering costs and boosting efficiency, growing the scale of agricultural land operations can strengthen the agricultural economic system’s resistance to market shocks. Second, the return of people of rural areas will support agricultural scientific and technological innovation, facilitate the efficient operation of large-scale land production, aid in the optimization and adjustment of the agricultural industrial structure, raise the value of agricultural products, and strengthen the agricultural economic system’s resilience to external risks. Simultaneously, the expansion of agricultural land management scale will allow agricultural business entities to change their operating methods, improve their own technical levels and management capabilities, and help the agricultural economic system evolve toward modernization.

2.2.2. Labor Efficiency Effect

Agricultural production efficiency is a measure of labor efficiency. Due to the fact that the majority of the population still moves from rural to urban areas, the decrease in the population in rural areas has resulted in higher labor costs for agriculture. Agricultural production has progressively moved toward mechanization in an effort to lower costs. The quantity of agricultural machinery has significantly increased in conjunction with the implementation of policies for agricultural machinery subsidies. Grain production can be increased and agricultural production efficiency raised through the use of agricultural machinery. Moreover, an increase in grain production can raise farmers’ operating income, which enables the agricultural economy to withstand market shocks to some extent. A strong economic base also offers financial support for the agricultural economic system’s recovery.

The following theories are put forth in light of the above theoretical analysis:

Hypothesis H2:

By improving agricultural production conditions, the bidirectional urban–rural mobility fosters agricultural economic resilience.

2.3. The Moderation Role of Smart Agriculture Development in the Bidirectional Flow of Urban and Rural Population and the Resilience of the Agricultural Economy

Marx’s “Technology-Economy” theory holds that economic growth is significantly influenced by technological advancement. It is essential for raising labor productivity, encouraging the modernization of industrial structures, and fostering economic expansion. Enhancing agricultural economic resilience will also be facilitated by the advancement of agricultural science and technology. The widespread use of smart agriculture in agricultural production has aided in the growth of the agricultural economy in recent years, particularly with the implementation of the rural revitalization strategy and the rapid development of digital villages. On the one hand, in order to monitor soil moisture, fertility, and other farmland data in real time, the smart agricultural system makes use of remote sensing drones, the internet of things, and other technologies. This enhances the agricultural economic system’s ability to anticipate and detect agricultural risks beforehand, which can increase its resilience. The efficiency of agricultural production has significantly increased with the use of intelligent platforms to guide drones to patrol fields and operate. Effective agricultural production efficiency can rapidly restore agricultural output and increase the resilience of the agricultural economic system following risks to it. On the other hand, the advent of smart agriculture has introduced cutting-edge science and technology to agricultural production, which can help the agricultural economy become more modern and reconstructive. Broadly implementing smart agriculture can also lessen the need for agricultural labor, partially address the labor shortage issue, and decrease the negative effects of rural population decline on the agricultural economy’s resilience.

The following theories are put forth in light of the above theoretical analysis:

Hypothesis H3:

As smart agriculture advances, the negative impact of both urban and rural population flow on agricultural economic resilience will be lessened.

3. Models

3.1. Model Construction

3.1.1. Benchmark Regression Model

This study builds the following benchmark regression model to examine the effects of bidirectional urban and rural population flow on agricultural economic resilience:

Aer_i,t = α₀ + α₁X_i,t + α₂control_i,t + μ_i + e_t + ε_i,t

(1)

where α₀ is a constant term, α₁ and α₂ are coefficients, while i and t stand for provinces and years, respectively. The dependent variable representing agricultural economic resilience is Aer. The explanatory variable for the bidirectional population flow between urban and rural areas is X. The random error term is denoted by ε. This paper also fixes the control variable (control_i,t), the fixed effect of province (fixed effect μ_i), and the fixed effect of time (fixed effect e_t) in consideration of the potential missing variables in the model.

3.1.2. Mediating Effect Model

Based on the benchmark regression model and with reference to the methods of Wen Zhonglin and Ye Baojuan [23], the following mediation effect model is constructed in order to further investigate the mechanism by which the bidirectional flow of urban and rural population affects agricultural economic resilience:

M_i,t = β₀ + β₁X_i,t + β₂control_i,t + μ_i + e_t + ε_i,t

(2)

Aer_i,t = γ₀ + γ₁X_i,t + γ₂M_i,t + γ₃control_i,t + μ_i + e_t + ε_i,t

(3)

where M is the intermediary variable, β₀ and γ₀ are constant terms; β₁, γ₁, β₂, γ₂, and γ₃ are coefficients; the meaning of other variables is the same as Formula (1).

3.2. Measurements

3.2.1. Explained Variable

This paper selects agricultural economic resilience (Aer) as the dependent variable. The agricultural economic resilience evaluation index system is built from three dimensions: resistance, restoration, and reconstruction, in accordance with the definition of agricultural economic resilience [14] and with reference to the methods of Hao Aimin and Tan Jiayin [24] and Zhao Wei et al. [25]. Development resilience, ecological resilience, and production resilience are all forms of resistance. Recovery resilience measures restoration, innovation resilience measures reconstruction, and its value is measured by the entropy method. The evaluation indicator system is shown in

Table 1. Agricultural economic resilience evaluation indicator system.

First-Level Indicators	Secondary Indicators	Third-Level Indicators	Units	Attribute	Weight
Resistance	Production resilience	Crop irrigated area/crop sown area	%	P	0.0314812
		Total power of agricultural machinery per unit sowing area	kW/ha	P	0.0383763
		Disaster-stricken area/affected area	%	N	0.0093801
		Agricultural meteorological stations	a	P	0.0193356
	Ecological resilience	Water consumption for agricultural production per unit of sown area	cubic meters/hectare	N	0.0066799
		Amount of agricultural chemical fertilizers (converted to pure) per unit sowing area	t/ha	N	0.0087877
		Agricultural diesel oil application per unit sowing area	kg/ha	N	0.0889373
		Pesticide application per unit sowing area	kg/ha	N	0.025624
	Development tenacity	Added value of agriculture, forestry, animal husbandry, and fishery/number of employees in agriculture, forestry, animal husbandry, and fishery	Yuan/person	P	0.0910319
		Added value of agriculture, forestry, animal husbandry, and fishery/crop sown area	100 million yuan/thousand hectares	P	0.0232981
		Per capita food production	Kilogram/person	P	0.0451883
Resilience	Restoring resilience	Fiscal expenditure on agriculture, forestry, and water/number of employees in agriculture, forestry, animal husbandry, and fishery	Ten thousand yuan/person	P	0.2078624
		Number of employees in agriculture, forestry, animal husbandry, and fishery/number of employees in rural employment	%	P	0.0985632
		Fiscal expenditure on agriculture, forestry, and water/total fiscal expenditure	%	P	0.0168658
Reconstruction	Innovation resilience	Per capita electricity consumption in rural areas	Ten thousand kilowatt-hours/person	P	0.1178231
		Farmland water-saving irrigation area/crop sown area	%	P	0.0987617
		Number of applications for new agricultural plant varieties	a	P	0.0720034

P represents positive impact; N represents negative impact (this was constructed by the authors).

.

3.2.2. Explanatory Variable

This paper selects the urban and rural migrant population (Migration) as the explanatory variable. In reference to the methodology of Li Fuyou et al. [26], this article decides to measure the bidirectional flow of the urban and rural populations using the rural population chain ratio (Rpcr) and the urban population chain ratio (Upcr). This type of measurement technique can show the flow between urban and rural populations as well as the growth trend of both urban and rural populations, in addition to reflecting the growth rate of both populations to a certain degree.

3.2.3. Mediating Variables

The level of agricultural production serves as the intermediary variable in this paper, and it is examined from two angles: land scale and labor efficiency. Among these, the land scale is determined by measuring the land area of large-scale agricultural operations (Land); labor efficiency is determined by measuring agricultural production efficiency (Ape), and the data for analysis is per capita grain output, in accordance with the practice of Jiang Jian et al. [14].

3.2.4. Control Variables

The following elements are chosen as control variables based on the research conducted for this paper and the body of existing literature: (1) The percentage of different crops sown is used to calculate the grain planting structure (St). (2) The degree of population aging (Poe) which refers to the practice of Yao Dongmin et al. [27] is measured by the proportion of the rural elderly population over 65 years old, given that our nation’s farmers have a long working life. (3) The quantity of rural hydropower station is a measure of agricultural infrastructure (Reservoir). (4) Ecological environment (Sec), which is determined by the area under soil erosion control. (5) The total technological power of agricultural machinery (Tech).

3.2.5. Moderating Variable

The number of agricultural aircraft serve as a proxy for the development status of smart agriculture (Sa), which is the moderator variable in this study.

3.3. Data

This study uses panel data from 31 Chinese provinces (not including Hong Kong, Macao, and Taiwan) from 2017 to 2022 because the idea of “rural revitalization” was first put forth in that year, and digital countryside also advanced quickly during that time. These include the interpreted variable agricultural economic resilience data from “China Rural Statistical Yearbook,” “China Science and Technology Statistical Yearbook,” “China Population and Employment Yearbook,” and the official website of the National Bureau of Statistics; the explanatory variable bidirectional flow data of urban and rural populations comes from “China Statistical Yearbook,” “China Urban and Rural Construction Statistical Yearbook,” and the National Bureau of Statistics’ official website. Additional information is derived from the “China Agricultural Machinery Industry Yearbook”, Guotai ‘an Database, and the National Bureau of Statistics’ official website.

The findings of the descriptive statistical analysis are displayed in

Table 2. Descriptive statistics.

Variable Type	Symbol	Variables	Sample Size	Average Number	Median	Standard Deviation	Minimum Value	Maximum Value
Explained variable	Aer	Agricultural economic resilience	186	0.209	0.199	0.0470	0.118	0.421
Explanatory variables	Rpcr	Rural population chain ratio	186	−0.0240	−0.0260	0.0120	−0.0610	0.00900
	Upcr	Urban population chain ratio	186	0.0200	0.0200	0.0160	−0.0200	0.0840
Control variables	St	Grain planting structure	186	0.650	0.644	0.156	0.355	0.971
	Poe	Degree of aging population	186	0.125	0.124	0.0300	0.0570	0.200
	Reservoir	Reservoir agricultural infrastructure	186	301.2	207	283.1	0	1264
	Sec	Ecological environment	186	4534	3817	3591	0	16,679
	Tech	Total power of agricultural machinery	186	3365	2539	2825	93.97	11,530
Mediating variables	Land	Land area for large-scale agricultural operations	186	4028	3756	3121	11.04	14,351
	Ape	Labor productivity	186	497.7	421.0	478.3	13.13	2499
Moderating variable	Sa	Development status of smart agriculture	185	2010	833	3051	0	19,374

This was constructed by the authors.

. Some variables’ sample sizes for descriptive statistical analysis were inconsistent because of missing data.

4. Empirical Results

4.1. Benchmark Regression Analysis

4.1.1. Correlation Analysis

The explained variables, explanatory variables, and control variables for benchmark regression analysis are first subjected to correlation analysis. The results of the correlation analysis show that the explained variable, agricultural economic resilience, has a strong positive correlation with the explanatory variable, the rural population chain relative ratio, and a negative correlation with the urban population chain relative ratio. This is because the larger the correlation coefficient, the stronger the correlation between the variables (

Table 3. Correlations among rural–urban mobility, agricultural economic resilience, and controls.

	Aer	Rpcr	Upcr	St	Poe	Reservoir	Sec	Tech
Aer	1
Rpcr	0.1640	1
Upcr	−0.4070	−0.0453	1
St	0.0396	−0.1360	−0.1450	1
Poe	0.2030	−0.0811	−0.5280	0.1990	1
Reservoir	−0.2190	−0.0738	0.1010	−0.0855	0.0457	1
Sec	−0.0825	−0.2640	0.0333	0.1530	0.1500	0.3130	1
Tech	0.0760	−0.2370	0.0042	0.1560	0.3480	0.2560	0.2170	1

This was constructed by the authors.

).

4.1.2. Collinearity Diagnosis

In general, multicollinearity is indicated by a VIF value larger than 10, which is the criterion for diagnosing collinearity. The results of the collinearity diagnosis in this paper indicate that the VIF values for the collinearity diagnosis are all less than 10, with the mean VIF being 2.510, the maximum value being 7.110 for Land, and the minimum value being 1.130 for St. As a result, the variables examined in this paper do not clearly exhibit multicollinearity.

4.1.3. Regression Analysis

The relationship between the explanatory variable and the explained variable is examined using regression analysis in this paper. The outcome of benchmark regression between the explained and explanatory variables using the least squares method is shown in column (1) of

Table 4. Benchmark regression.

	(1)	(2)	(3)
	OLS	Bidirectional Fixed Model	Tobit
	Aer	Aer	Aer
Main
Rpcr	0.666 **	0.808 ***	0.666 **
	(0.271)	(0.266)	(0.265)
Upcr	−1.224 ***	−1.188 ***	−1.224 ***
	(0.237)	(0.218)	(0.231)
St	−0.002	−0.001	−0.002
	(0.003)	(0.003)	(0.003)
Poe	−0.003	−0.006	−0.003
	(0.004)	(0.004)	(0.004)
Reservoir	−0.010 ***	−0.011 ***	−0.010 ***
	(0.003)	(0.003)	(0.003)
Sec	0.001	0.001	0.001
	(0.003)	(0.003)	(0.003)
Tech	0.009 **	0.010 ***	0.009 ***
	(0.004)	(0.003)	(0.003)
_cons	0.249 ***	0.252 ***	0.249 ***
	(0.008)	(0.009)	(0.008)
/
var (e.Aer)			0.002 ***
			(0.000)
N	186.000	186.000	186.000
r2	0.245	0.256
r2_a	0.216	0.204

Standard errors in parentheses. ** p < 0.05, *** p < 0.01 (this was constructed by the authors).

. The outcome of regression using a bidirectional fixed model is shown in column (2). The results are displayed in column (3). The restricted variable model (Tobit) is used to re-estimate because the value range of the explained variable agricultural economic resilience in this paper is [0, 1], which satisfies the requirements of the restricted dependent variable model. According to the results, there is a significant positive correlation between agricultural economic resilience and rural population chain relative ratio, while at the 1% level, the coefficient of urban population chain relative ratio is significantly negative. This indicates that while an increase in urban population chain relative ratio will hinder agricultural economic resilience, an increase in rural population chain relative ratio will foster it. Hypothesis H1 has been confirmed since the effect of bidirectional urban and rural population flow on agricultural economic resilience is inhibited because the coefficient of urban population chain relative ratio is substantially larger than the coefficient of rural population chain relative ratio.

4.1.4. Endogeneity Test

The 2SLS regression results indicate that terrain conditions (Land) are significantly positively correlated with the rural population year-on-year change rate (Rpcr), while terrain conditions (Land) are not significantly correlated with agricultural economic resilience. This demonstrates the validity of selecting this instrumental variable. Furthermore, the Two-Stage Least Squares (2SLS) method effectively resolved the endogeneity issue present in the original regression. The results are shown in

Table 5. Endogeneity test results.

	(1)	(2)
	First	Second
Variables	Rpcr	Aer
Land	0.0001 ***
	(3.41)
Rpcr		3.2266 **
		(2.35)
St	−0.0054	0.0109
	(−0.98)	(0.41)
Poe	0.0618 *	0.2613 *
	(1.91)	(1.90)
Reservoir	0.0000	−0.0000 ***
	(0.79)	(−2.98)
Sec	−0.0000 **	0.0000
	(−2.54)	(1.13)
Tech	−0.0000 **	0.0000 **
	(−2.37)	(2.15)
Constant	−0.0248 ***	0.2382 ***
	(−4.73)	(7.64)
Observations	186	186
R-squared		−0.229

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01 (this was constructed by the authors).

.

4.2. Robustness Tests

The robustness of the relationship between the explanatory variable, the bidirectional flow of urban and rural population, and the explained variable, agricultural economic resilience, is tested in this paper using the following four techniques. The results are shown in

Table 6. Robustness tests.

	(1)	(2)	(3)	(4)
	Replace Explanatory Variables	Replace Control Variables	Sample Change Time	Quantile Regression
	Aer	Aer	Aer	Aer
Main
rpgr	0.793 ***
	(0.287)
Rpcr		0.831 **	0.536 *	1.073 ***
		(0.409)	(0.311)	(0.409)
Upcr		−1.328 ***	−1.473 ***	−1.386 ***
		(0.270)	(0.313)	(0.357)
St	−0.001		−0.001	−0.009 *
	(0.003)		(0.004)	(0.005)
Poe	0.009 **		−0.005	−0.003
	(0.004)		(0.005)	(0.006)
Reservoir	−0.011 ***		−0.010 ***	−0.013 **
	(0.004)		(0.004)	(0.005)
Sec	−0.000		0.000	0.003
	(0.004)		(0.004)	(0.005)
Tech	0.006		0.008 **	0.009 *
	(0.004)		(0.004)	(0.005)
Apf		0.005
		(0.005)
Hhc		0.002
		(0.005)
Capital		−0.009 **
		(0.004)
Pcec		0.000
		(0.003)
Gap		−0.002
		(0.005)
_cons	0.228 ***	0.259 ***	0.252 ***	0.283 ***
	(0.008)	(0.012)	(0.010)	(0.013)
N	186.000	129.000	155.000	186.000
r2	0.138	0.251	0.221
r2_a	0.109	0.208	0.184

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01 (this was constructed by the authors).

.

The first way to start a robustness test is to substitute rural population growth rate (rpgr) for the explanatory variable and run a regression analysis. In line with the regression of rural population chain relative ratio on agricultural economic resilience, the regression result column (1) shows that the relationship between the explanatory variable agricultural economic resilience and the substituted explanatory variable rural population growth rate is significantly positive at the 1% level.

The second approach, which is based on the work of Li Tong et al. [28], substitutes agricultural plastic films (apf), health human capital (hhc), agricultural farmers’ capital investment (capital), per capita electricity consumption (pcec), and total output value of agriculture (gap) for the control variables. A regression analysis is then conducted between the explanatory variables, rural population chain relative ratio and urban population chain relative ratio, and the explained variable agricultural economic resilience, as indicated in Table 5, column (2). The regression results demonstrate that there is little difference between the initial regression result and the regression result following the replacement of the control variables. The positive and negative traits of the agricultural economic resilience coefficients are comparable to those of the rural and urban populations.

The third approach makes reference to the study of Xue Qiutong et al. [29], which chooses data from 2018 to 2022 as a sample for regression and performs a robustness test by varying the sample time. In column (3), the regression results are displayed.

Li Min’s method is referred to in the fourth method [30]. The relationship between the explained variable and the explanatory variable at the 0.75 quantile is examined using quantile regression in column (4). Regression analysis reveals that the explanatory variable rural population chain relative ratio has a significantly positive regression result with the explained variable agricultural economic resilience, while the explanatory variable urban population chain relative ratio has a significantly negative regression result with the explained variable agricultural economic resilience. The urban population chain relative ratio coefficient’s absolute value is higher than the rural population chain relative ratio coefficient’s, and the robustness of H1 is tested.

4.3. Mechanism Analysis

4.3.1. The Mediating Effect of the Level of Agricultural Production

The degree of agricultural production, as determined by the land area of large-scale agricultural operations (Land) and agricultural production efficiency (Ape), serves as the mediating variable in this study.

When we combine column (1), we can see that the rural population chain relative ratio promotes agricultural economic resilience by suppressing the land scale effect, while the urban population chain relative ratio promotes agricultural economic resilience by promoting the land scale effect. This is evident from columns (2) and (3) of

Table 7. Analysis of mediation effects.

	(1)	(2)	(3)	(4)	(5)	(6)
	Aer	Land	Aer	Aer	Ape	Aer
Rpcr	0.170 **	−0.188 ***	0.216 ***	0.170 **	−0.265 ***	0.244 ***
	(0.069)	(0.042)	(0.072)	(0.069)	(0.059)	(0.071)
Upcr	−0.413 ***	−0.182 ***	−0.368 ***	−0.413 ***	−0.420 ***	−0.295 ***
	(0.080)	(0.048)	(0.082)	(0.080)	(0.068)	(0.086)
Land			0.245 **
			(0.124)
Ape						0.281 ***
						(0.086)
St	−0.038	−0.037	−0.029	−0.038	−0.029	−0.030
	(0.068)	(0.041)	(0.068)	(0.068)	(0.058)	(0.067)
Poe	−0.054	−0.061	−0.039	−0.054	−0.221 ***	0.008
	(0.085)	(0.051)	(0.085)	(0.085)	(0.072)	(0.085)
Reservoir	−0.221 ***	0.055	−0.235 ***	−0.221 ***	−0.169 ***	−0.174 **
	(0.071)	(0.043)	(0.071)	(0.071)	(0.061)	(0.071)
Sec	0.017	0.129 ***	−0.015	0.017	0.283 ***	−0.063
	(0.072)	(0.044)	(0.074)	(0.072)	(0.062)	(0.075)
Tech	0.196 **	0.745 ***	0.013	0.196 **	0.424 ***	0.077
	(0.075)	(0.045)	(0.119)	(0.075)	(0.064)	(0.082)
_cons	−0.000	0.000	−0.000	−0.000	−0.000	−0.000
	(0.065)	(0.039)	(0.064)	(0.065)	(0.055)	(0.063)
N	186.000	186.000	186.000	186.000	186.000	186.000
r2	0.245	0.727	0.262	0.245	0.453	0.288
r2_a	0.216	0.716	0.228	0.216	0.432	0.256

Standard errors in parentheses. ** p < 0.05, *** p < 0.01 (this was constructed by the authors).

, which show that the regression result of rural population chain relative ratio on the land area of large-scale agricultural operation is significantly negative at the 1% level, and the regression result of urban population chain relative ratio on the land area of large-scale agricultural operation is significantly negative at the 1% level. When rural populations leave, the outflow population typically leases land to these entities. Large-scale agricultural operation entities typically choose to operate land through agricultural mechanization or automation, which lowers agricultural operation costs to some extent and improves agriculture’s ability to withstand risks, which in turn fosters the resilience of the agricultural economy.

Both the regression results of the urban population chain relative ratio on agricultural production efficiency are at the 1% level, and the regression results of the rural population chain relative ratio on agricultural production efficiency are significantly negative at the 1% level, and the regression results of agricultural production efficiency on agricultural economic resilience are significantly positive at the 1% level, as shown in columns (5) and (6) of Table 7. Combining column (1), we can observe that the rural population chain relative ratio promotes agricultural economic resilience by suppressing labor production benefits, while the urban population chain relative ratio promotes agricultural economic resilience by promoting labor production benefits. Due to the shortage of agricultural labor caused by the rural population’s exodus, agricultural operators are urging reasonable labor distribution and increased agricultural production efficiency through prudent land transfer and capital investment, which will strengthen the agricultural economy’s resilience. This led to the verification of hypothesis H2.

4.3.2. The Moderating Effect of the Development Status of Smart Agriculture

This paper uses the number of agricultural aircraft to determine the state of smart agriculture development in the context of digital villages (Sa) as an adjustment variable.

The benchmark regression results for the explained and explanatory variables are shown in column (1) of

Table 8. Analysis of regulatory effects.

	(1)	(2)	(3)	(4)
	Aer	Aer	Aer	Aer
Rpgb	0.145 **	0.170 **	0.207 ***	0.219 ***
	(0.067)	(0.069)	(0.069)	(0.072)
Upgb	−0.401 ***	−0.413 ***	−0.318 ***	−0.328 ***
	(0.067)	(0.080)	(0.074)	(0.087)
Rpgb × Sa			−0.303 ***	−0.262 **
			(0.091)	(0.101)
Upgb × Sa			−0.163 *	−0.173 *
			(0.088)	(0.092)
Controls	No	Yes	No	Yes
_cons	−0.000	−0.000	−0.009	−0.009
	(0.066)	(0.065)	(0.064)	(0.064)
N	186.000	186.000	185.000	185.000
r2	0.187	0.245	0.246	0.284
r2_a	0.178	0.216	0.229	0.247

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01 (this was constructed by the authors).

, while the regression results for the explained variable and the interaction term that results from multiplying the explanatory and adjusting variables are shown in column (3). When comparing the regression results of columns (2) and (4), we can see that the interaction between the rural population chain relative ratio and the development status of smart agriculture has a significantly negative impact on agricultural economic resilience, indicating that the development of smart agriculture will lessen the positive impact of the rural population chain relative ratio on agricultural economic resilience. The regression result between the urban population chain relative ratio and the development status of smart agriculture and agricultural economic resilience is also significantly negative, indicating that the development of smart agriculture will lessen the negative impact of the urban population chain relative ratio on agricultural economic resilience. In the context of digital villages, traditional planting no longer depends on the accumulation of labor due to the rapid advancement of automation and mechanization in smart agriculture. Thus, the development of smart agriculture will lessen the negative effects of the bidirectional flow of people between urban and rural areas on agricultural economic resilience. This led to the verification of hypothesis H3.

4.4. Heterogeneity Analysis

4.4.1. Heterogeneity Analysis of Production Conditions

To illustrate the most impactful production conditions, we selected the total volume of large agricultural machinery to measure mechanization and subsidies for agricultural machinery purchases (SAPs) to measure the implementation of agricultural development policies. We stratified the entire sample into four distinct groups based on these metrics—High Mechanization and High SAPs, High Mechanization and Low SAPs, Low Mechanization and High SAPs, and Low Mechanization and Low SAPs—for heterogeneity analysis. The results of this analysis are presented in the table below.

Table 9. Heterogeneity analysis of production conditions.

	(1)	(2)	(3)	(4)
	High Mechanization and High SAPs	High Mechanization and Low SAPs	Low Mechanization and High SAPs	Low Mechanization and Low SAPs
	Aer	Aer	Aer	Aer
Rpcr	−0.014	0.209	−0.136	0.287 **
	(0.102)	(0.442)	(0.470)	(0.113)
Upcr	−0.477 ***	0.115	0.271	−0.409 ***
	(0.126)	(0.452)	(0.372)	(0.131)
St	−0.018	0.823 **	−0.325	−0.203 **
	(0.112)	(0.335)	(0.766)	(0.100)
Poe	−0.167	0.390	1.083	0.031
	(0.163)	(0.481)	(0.547)	(0.120)
Reservoir	−0.281 ***	−0.554	0.637	−0.192 *
	(0.096)	(0.397)	(0.739)	(0.111)
Sec	0.143	−0.405	−0.967	0.111
	(0.129)	(0.373)	(0.568)	(0.101)
Tech	0.315 ***	0.195	−1.036	0.034
	(0.111)	(0.337)	(0.916)	(0.124)
_cons	−0.108	0.479	−0.259	−0.033
	(0.120)	(0.335)	(0.790)	(0.100)
N	57.000	15.000	12.000	96.000
r2	0.401	0.762	0.724	0.300
r2_a	0.315	0.525	0.242	0.245

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01 (this was constructed by the authors).

shows that the resilience of the agricultural economy is negatively impacted by the rural population chain relative ratio when it receives high subsidies, and that the negative impact of low mechanization is greater than the negative impact of high mechanization; conversely, when it receives low subsidies, the rural population chain relative ratio has a positive impact on the agricultural economy’s resilience, and the positive impact of low mechanization is greater than the negative impact of high mechanization. The growing rural population chain relative ratio will exacerbate labor idleness and the fall in marginal output at low mechanization levels, weakening economic resilience; excessive subsidies may distort resource allocation, forcing farmers to rely on outside funding rather than maximizing production efficiency; and mechanization partially mitigates the adverse effects of labor surplus at high levels of mechanization. Conversely, in situations with low subsidies, farmers must become more self-sufficient, and the rural population grows in order to supply inexpensive labor, encourage labor-intensive production, and boost resilience; in situations with low levels of mechanization, labor is the primary input, whereas in those with high levels of mechanization, machinery has increased efficiency, and the marginal benefits of extra labor are minimal. This illustrates how production technologies and subsidy policies interact: low subsidies promote efficient labor use and build resilience, while high subsidies may stifle innovation and increase population pressure.

The urban population chain relative ratio has a negative impact on the agricultural economic resilience when the degree of mechanization and the subsidies obtained are both high and low. The negative impact of high mechanization and high subsidies is greater than the negative impact of low-mechanization base subsidies. On the other hand, the urban population chain relative ratio has a positive impact on the agricultural economic resilience when the degree of mechanization and the subsidies obtained are both high and low. The positive effect of low mechanization and high subsidies is greater than the positive effect of high-mechanization base subsidies. High subsidies and high mechanization can easily cause agriculture to rely on capital investment and overlook manpower optimization, which is one of the potential causes. Urban population growth at this time will exacerbate the exodus of young, skilled workers from rural areas, creating the paradox of “excess capital but hollow manpower” and reducing the resilience of the system. Because capital-intensive systems are more susceptible to labor losses, the negative effect is more pronounced. Agriculture depends on traditional labor due to minimal mechanization and low subsidies. The resilience of the agricultural economy is diminished by urban expansion, which directly extracts labor as a fundamental factor of production and lacks capital and technical compensation. Subsidies, however, can compensate for technical deficiencies when there is a high level of subsidies and little mechanization. Increases in urban population stimulate idle labor, increase resilience, and result in the spillover of capital and technology. Furthermore, subsidies increase and elevate the value of labor. Mechanization preserves basic efficiency in the high mechanization and low subsidy scenarios, while low subsidies limit technological advancements. Urban expansion’s surplus capital can bridge the funding gap and provide some positive support, but the benefits are diminished because of the lack of technological iteration.

4.4.2. Resilience Dimension

This paper dissects the agricultural economic resilience according to its three dimensions in order to further validate H1. It is then tested using a separate regression with the bidirectional flow of the urban and rural populations. This study examines the relationship between the explanatory variable of bidirectional population flow between urban and rural areas and the explained variable of agricultural economic resilience from three perspectives: resistance, restoration, and reconstruction.

The regression results in

Table 10. Fractional regression analysis.

	(1)	(2)	(3)
	Res.	Rest.	Recons.
Rpcr	−2.432 ***	3.889 ***	1.221 **
	(0.378)	(0.837)	(0.513)
Upcr	0.127	−2.694 ***	−1.396 ***
	(0.330)	(0.730)	(0.447)
Controls	Yes	Yes	Yes
_cons	0.281 ***	0.294 ***	0.155 ***
	(0.012)	(0.026)	(0.016)
N	186.000	186.000	186.000
r2	0.437	0.290	0.245
r2_a	0.415	0.262	0.216

Standard errors in parentheses. ** p < 0.05, *** p < 0.01 (this was constructed by the authors).

’s columns (1) through (3) show that there is a significant negative correlation between the rural population chain relative ratio and the resistance of the agricultural economic resilience, while there is a significant positive correlation between the rural population chain relative ratio and the restoration and reconstruction of the agricultural economic resilience. At the 1% level, there is a significant negative correlation between the urban population chain relative ratio and the restoration and reconstruction of the agricultural economic resilience. Among the potential causes are the following: The chain relative ratio increase in the rural population suggests either a slower rate of decline or a faster rate of growth. Given that the rural population chain relative ratio is primarily negative, the chain relative ratio increase in the rural population indicates a slower rate of decline in the rural population; in other words, the rural environment has not improved to a certain extent, which makes it harder for the countryside to withstand agricultural risks. On the other hand, the return of rural residents to cities contributes a wealth of human resources to rural development and can improve the ability of rural areas to restore damaged agricultural systems, simultaneously bringing new agricultural technologies to rural areas and promoting agricultural renewal. Table 10 shows how the rural population chain relative ratio affects the three aspects of agricultural economic resilience: resistance, restoration, and reconstruction. The development of resistance largely depends on the transmission of traditional agricultural knowledge and the growth of production scale diversity, meanwhile the development of restoration depends on the number of rural workers, and the development of reconstruction is more likely to be impacted by the advancement of agricultural science and technology.

4.4.3. Further Differentiation

This paper uses the month-on-month growth of the rural population (Mrpg) and the month-on-month growth of the urban population (Mupg) to measure the reverse flow of urban and rural population in order to further analyze the effects of bidirectional flow of urban and rural population, rural population return, and urban population flow to rural areas on the resilience of the agricultural economy. Analysis is carried out on the correlation between the three aspects of resistance, restoration, and reconstruction of the agricultural economic resilience.

Table 11. Further dimensional regression analysis.

	(1)	(2)	(3)
	Res.	Rest.	Recons.
Mrpg	0.026	−0.060	0.029
	(0.063)	(0.069)	(0.063)
Mupg	−0.007	−0.180 **	0.260 ***
	(0.065)	(0.070)	(0.064)
St	0.046	−0.009	−0.079
	(0.066)	(0.071)	(0.066)
Poe	−0.533 ***	0.338 ***	0.302 ***
	(0.071)	(0.076)	(0.070)
Reservoir	−0.148 **	−0.070	−0.129*
	(0.069)	(0.074)	(0.068)
Sec	0.073	0.043	−0.256 ***
	(0.068)	(0.074)	(0.068)
Tech	0.424 ***	−0.235 ***	0.124 *
	(0.071)	(0.076)	(0.070)
_cons	−0.007	0.012	−0.017
	(0.063)	(0.068)	(0.062)
N	184.000	184.000	184.000
r2	0.305	0.189	0.241
r2_a	0.277	0.157	0.211

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01 (this was constructed by the authors).

displays the findings. The natural population growth rate in both urban and rural areas is assumed to be zero in the regression analysis, meaning that the birth and death rates stay constant. According to a detailed examination of the data, the rural population chain relative ratio is generally negative, meaning that the rural population as a whole is on the decline, while the growth of the rural population chain relative ratio is generally positive, meaning that the decline is becoming less pronounced; the urban population chain relative ratio is the opposite. Since the natural population growth rate is assumed to be zero, the only plausible explanation for this trend is the migration of urban residents to rural areas, that is, the reversal of the urban–rural population flow.

Consequently, the month-on-month growth of rural population has no discernible effect on resistance, restoration, and reconstruction when paired with the regression results in Table 11. This could be due to mutual hedging between the inflow and outflow of people to the countryside, even though there may be a comparison increase or decrease in the rural population chain relative ratio over a given time period. Although some rural labor has moved into cities, the quality of the labor force remaining in rural areas to work in agriculture is gradually improving due to rising educational standards and the popularity of vocational training. This could be the reason why the month-on-month growth of the urban population has no discernible effect on resistance. Professional agricultural technicians and managers, returning business owners, and some excellent farmers can use cutting-edge technology and management expertise to carry out agricultural production and enhance its efficiency and resistance to risk. To a certain degree, this counteracts the detrimental effect of urban population growth on the decline in agricultural labor. As a result, there is no meaningful connection between the two. It is possible that the reverse flow of urban and rural population means that rural areas have more labor, allowing agricultural production to have enough labor, which is conducive to a faster recovery of the agricultural economy after encountering risks. This is why the month-on-month growth of the urban population has a significantly negative impact on restoration; the demand for urban construction land’s continued expansion, the large-scale conversion of high-quality cultivated land into urban land, and the declining land resources for agricultural production could all be contributing factors that the urban population’s continued growth affect the reconstruction of the agricultural economic resilience. These factors not only directly limit the large-scale and intensive development of agriculture, but also make it more difficult to quickly adapt its production layout and scale up after experiencing disasters or market fluctuations.

4.4.4. Regional Decomposition

Thirty-one Chinese provinces are separated into eastern, central, and western regions based on geographic location. These regions are further divided into main grain-producing areas, balanced production and marketing areas, and main grain-selling areas based on geographical environment, agricultural foundation, and quality. Corresponding empirical tests are conducted to further investigate whether the impact of bidirectional urban and rural population flow on agricultural economic resilience is consistent across different regions.

The verification results from the eastern, central, and western regions are displayed in

Table 12. Regression analysis by province from the perspectives of east, middle, and west.

	East			Middle			West
	Res.	Rest.	Recons.	Res.	Rest.	Recons.	Res.	Rest.	Recons.
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Rpcr	−1.791 ***	4.471 **	0.824	−2.943 ***	0.199	0.051	−1.579 **	3.080 **	−0.249
	(0.667)	(1.830)	(1.344)	(0.875)	(1.327)	(0.653)	(0.615)	(1.334)	(0.323)
Upcr	−0.968	−0.727	−1.707	−0.159	−1.368	−0.889	1.957 ***	−5.132 ***	−0.342
	(0.670)	(1.838)	(1.349)	(0.769)	(1.165)	(0.574)	(0.479)	(1.039)	(0.252)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
_cons	0.275 ***	0.248 ***	0.209 ***	0.290 ***	0.096	0.063 *	0.286 ***	0.376 ***	0.073 ***
	(0.023)	(0.063)	(0.046)	(0.042)	(0.063)	(0.031)	(0.022)	(0.047)	(0.011)
N	66.000	66.000	66.000	54.000	54.000	54.000	66.000	66.000	66.000
r2	0.490	0.405	0.179	0.522	0.658	0.582	0.646	0.450	0.394
r2_a	0.429	0.333	0.080	0.449	0.606	0.518	0.604	0.384	0.321

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01 (this was constructed by the authors).

. It can be determined that the impact of the rural population chain relative ratio in the eastern region is significantly positive by combining the coefficients with the weights of resistance, restoration, and reconstruction in measuring agricultural economic resilience. Similarly, the rural population has a significantly negative impact on the agricultural economic resilience in the central region and a significantly positive impact on the agricultural economic resilience in the western region. One possible explanation is that the central region’s rural economic development depends more on planting than the eastern and western regions do. In order to develop rural planting, labor is needed. In the east and west, an increase in the rural population may bring more labor resources, as evidenced by the increase in the rural population chain relative ratio, which shows that the rural population loss is declining. To encourage the scale and modernization of agricultural operations, some of this labor force will remain in agriculture even though some may move to non-agricultural industries. Furthermore, there are a lot of undeveloped land resources in the western region. Increased labor investment in agricultural reclamation and production may result from the comparative growth in the rural population, which would help to increase the scale of agricultural production. The urban population chain relative ratio has a significantly negative impact on the agricultural economic resilience in the eastern and central regions, with the eastern region experiencing a greater negative impact than the central region. In contrast, the western region experiences a significantly positive impact on the agricultural economic resilience. This may make sense given that the eastern region is economically developed, urbanization is happening quickly, the population of cities is expanding quickly, and there is a high demand for land for urban growth. The conversion of a significant portion of prime agricultural land to urban construction has significantly decreased the amount of land available for agricultural production, made it more challenging to sustain and grow large-scale farming operations, severely undermined the agricultural economy’s foundation, and decreased the agricultural economic resilience. Despite facing the same issues as the eastern region, the central region has historically been a significant agricultural producing area in our nation, and as a result, the resilience of its agricultural economy has declined less than that of the eastern region. Nonetheless, the western region’s urban development is comparatively lagging behind. To a certain degree, urban population growth can contribute more resources, including money, technology, and skills, to the advancement of agriculture. The growth of cities can support agriculture and the processing and distribution of agricultural products and increase the agricultural economic resilience by offering a broader-ranging market and more comprehensive industrial chain support.

The verification results are displayed in

Table 13. Regression analysis by province from the perspective of grain balance.

	Main Grain-Producing Areas			Balanced Production and Marketing Areas			Main Grain-Selling Areas
	Res.	Rest.	Recons.	Res.	Rest.	Recons.	Res.	Rest.	Recons.
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Rpcr	−2.828 ***	0.460	−0.356	−1.546 **	3.113 **	−0.353	−1.820 **	5.010 *	0.744
	(0.643)	(1.056)	(0.550)	(0.616)	(1.420)	(0.328)	(0.819)	(2.611)	(1.903)
Upcr	0.159	−2.965 ***	−0.617	1.770 ***	−5.057 ***	−0.240	−2.088 **	1.601	−3.984 *
	(0.484)	(0.795)	(0.414)	(0.472)	(1.089)	(0.251)	(0.848)	(2.703)	(1.970)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
_cons	0.291 ***	0.138 ***	0.057 **	0.254 ***	0.348 ***	0.069 ***	0.096	0.464	−0.267
	(0.028)	(0.046)	(0.024)	(0.019)	(0.045)	(0.010)	(0.094)	(0.300)	(0.219)
N	78.000	78.000	78.000	66.000	66.000	66.000	42.000	42.000	42.000
r2	0.515	0.560	0.566	0.653	0.412	0.391	0.526	0.475	0.279
r2_a	0.467	0.516	0.523	0.611	0.341	0.317	0.428	0.367	0.131

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01 (this was constructed by the authors.).

from the viewpoints of the primary grain-producing regions, the balanced production and marketing regions, and the primary grain-selling regions. When combined with the coefficient and resistance, restoration, and reconstruction weights for assessing the agricultural economic resilience, it is evident that the rural population has a significantly negative comparison impact on the primary grain-producing regions and the balance of production and marketing; it has an adverse effect in the primary grain-producing regions more than on the balance of production and marketing regions; and there is a notable improvement in the agricultural economic resilience of the primary grain-selling regions. The primary grain-producing regions are the backbone of my nation’s grain production and have a high labor demand, which could be the cause. A significant amount of agricultural labor is being lost as a result of the comparative decline in the rural population. Additionally, it is challenging to concentrate land in the hands of large professional households or agricultural enterprises for large-scale operations because of the pervasive decentralization of land in major grain-producing areas and the flawed land transfer mechanism. In addition to decreasing agricultural production efficiency, this also raises agricultural production costs and erodes the agricultural economy’s resilience. In addition to meeting the local demand for food, the agricultural industry structure in the production and marketing balance area is rather complex, requiring the development of cash crop planting and distinctive agriculture. Both the quantity and quality of the agricultural work force have been impacted by the rural population drop. Lack of adequate technical and human support during the industrial structural adjustment process makes it challenging to quickly adjust to changes in the market. This leads to the agricultural economic inability to withstand external shocks, which in turn impacts the resistance of agricultural economic resilience. Comparatively speaking, the urban population has a significantly negative impact on the agricultural economic resilience of main grain-producing areas, balanced production and marketing areas, and main grain-selling areas. The negative impact is arranged as follows: The balanced production and marketing area is smaller than the main grain-producing area, which is smaller than the main grain-selling area. The primary grain-selling regions are often economically developed, cities are growing quickly, and the comparative increase in the urban population creates a high demand for land, all of which could be contributing factors. The foundation of agricultural production has been severely damaged, and the resilience of the agricultural economy has been significantly impacted as a result of the rapid conversion of a large portion of high-quality agricultural land into urban construction land; the primary grain-producing regions are heavily dependent on the market and produce a lot of agricultural goods. Grain production is the primary basis of the comparatively single agricultural production structure found in the major grain-producing regions. Farmers and agricultural business entities have little bargaining power in a volatile market, which makes it challenging to swiftly modify production plans and sales tactics. They are also more susceptible to changes in market prices, which weakens the agricultural economy’s resilience; the proportion of agriculture in the regional economy is relatively small, factors affecting agricultural production are comparatively less appealing to the growth of urban population chain relative ratio, and the urban population chain relative ratio is comparatively weak in relation to agricultural production factors. There is comparatively little loss of rural labor in the production and marketing balanced area when compared to the main grain-producing and main grain-selling areas, and cities occupy a comparatively small amount of land in such regions. As a result, the negative influence exerted on agricultural economic resilience is minimal and the basis of agricultural production remains comparatively stable in these balanced areas.

To further verify the impact of technological progress in different regions on the relationship between bidirectional urban–rural population flow and agricultural economic resilience, this paper divides China’s 31 provinces into three regions, eastern, central, and western, and conducts separate moderating effect analyses. The results are shown in

Table 14. Regional heterogeneity analysis of moderating effects.

	(1)	(2)	(3)
	East	Central	West
	Aer	Aer	Aer
Rpcr	0.247	−0.178	0.059
	(0.156)	(0.123)	(0.088)
Upcr	−0.340	−0.014	−0.200 *
	(0.208)	(0.182)	(0.106)
Rpcr × Sa	−0.059	−0.223 **	−0.217
	(0.356)	(0.103)	(0.465)
Upcr × Sa	−0.163	−0.169 *	0.866 **
	(0.426)	(0.100)	(0.356)
St	0.371 **	−0.011	−0.046
	(0.163)	(0.095)	(0.074)
Poe	0.208	0.525 ***	−0.258 **
	(0.206)	(0.173)	(0.101)
Reservoir	−0.275	−0.146 *	−0.011
	(0.219)	(0.074)	(0.092)
Sec	0.001	0.449 ***	−0.044
	(0.383)	(0.099)	(0.134)
Tech	0.024	0.466 ***	0.014
	(0.185)	(0.117)	(0.398)
_cons	−0.136	−0.658 ***	0.160
	(0.261)	(0.189)	(0.222)
N	66.000	54.000	65.000
r2	0.414	0.688	0.521
r2_a	0.319	0.624	0.443

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01 (this was constructed by the authors).

.

According to the regression results in columns (1) to (3) of Table 14, the interaction term between the rural population year-on-year change rate (Rpcr) and the level of smart agriculture development has a negative effect on agricultural economic resilience. This indicates that the development of smart agriculture diminishes the positive impact of the rural population change rate on agricultural economic resilience. The magnitude of this negative effect follows this order: central region > western region > eastern region. The possible reasons are as follows: The central region faces the most intense contradiction between population loss and the promotion of smart agriculture, leading to the most significant negative effect of the interaction term. In the western region, the foundation for the “amplification” effect is weaker, partly because the depth and breadth of smart agriculture promotion itself may be less than in the central region. Additionally, facing enormous comprehensive challenges, the marginal benefit of the smart agriculture effect is relatively smaller. In the eastern region, when facing population loss, its relatively stronger economic foundation, talent pool, scale conditions, and infrastructure can partially offset the negative effect of the interaction term, making its negative impact relatively the smallest.

Simultaneously, the regression results for the interaction term between the urban population year-on-year change rate (Upcr) and the level of smart agriculture development on agricultural economic resilience are negative in the eastern and central regions, with the negative effect is larger in the central region than in the eastern region. However, in the western region, the effect is significantly positive. This indicates that in the eastern and central regions, smart agriculture development mitigates the negative impact of the urban population change rate on agricultural economic resilience. In the western region, smart agriculture development amplifies the negative impact of the urban population change rate on agricultural economic resilience. The possible reasons are that agriculture in the eastern region, due to economic prosperity, high non-agricultural income, widespread pluriactivity, and a more complete industrial chain, may have higher initial resilience to the shock of labor loss compared to the central region. Therefore, the marginal buffering effect of smart agriculture is less pronounced than in the more vulnerable central region. Simultaneously, as China’s primary grain-producing area, labor loss poses a systemic threat to agricultural production in the central region. Consequently, the stabilizing contribution to resilience from smart agriculture development, which effectively addresses this core challenge, is more significant in the central region. Many areas in the western region have not yet reached the mature stage of industrialization and urbanization seen in the east. Agriculture still accounts for a high proportion of the population and plays a vital role in livelihood security. Underpinned by a less solid foundation, the premature implementation of costly smart agriculture alongside rapid urbanization may trigger systemic imbalance. This can amplify the pains of transition and ultimately damage agricultural resilience instead.

5. Conclusions and Suggestion

Under the rapidly developing context of digital rural development, the bidirectional flow of people between urban and rural areas plays a vital role in promoting the sustainable development of agricultural economic resilience. This study employed panel data from 31 provincial-level regions in China spanning 2017–2022 to comprehensively examine the impact of bidirectional urban–rural mobility on diverse dimensions of agricultural economic resilience, while further investigating its underlying mechanisms. This study has three key conclusions: First, the bidirectional urban–rural mobility exerts a suppressive effect on the resistance and reconstruction of agricultural economic resilience, while concurrently enhancing its restoration. Second, such mobility contributes to strengthening agricultural economic resilience by catalyzing land-scale operational efficiency and amplifying labor productivity gains. Third, the advancement of smart agriculture technologies effectively mitigates the inhibitory impacts of bidirectional mobility on agricultural economic resilience. Furthermore, according to heterogeneity analysis, the bidirectional mobility of the urban and rural populations has significantly impeded the agricultural economic resilience of the eastern, central, and western regions, as well as the primary grain-producing areas, production and marketing balance areas, and the primary grain-selling areas. Further investigation reveals that the reverse mobility of the urban and rural populations has a positive effect on the agricultural economy’s resistance but a negative effect on its restoration and reconstruction. This study provides the following insights based on the aforementioned conclusions:

In order to promote urban migration to rural regions and support the quick development of digital villages, the state should first implement rural welfare programs. The trajectory of urban growth in my nation is now positive. The development of the rural economy should be vigorous in order to attain prosperity for all. As a result, the adoption of appropriate economic preferential policies can both support the growth of agricultural mechanization and encourage the construction of sophisticated communities in the modern period.

Second, in order to increase agricultural production levels, the state ought to encourage the extensive planting of rural land. The core of large-scale land operation lies in innovating land transfer mechanisms, strengthening the cultivation of new agricultural business entities, and improving the socialized service system. This involves establishing robust land transfer markets and service platforms to facilitate the consolidation of idle or inefficient land; providing targeted policy support (funding, technology, insurance) to foster new entities like family farms, cooperatives, and agricultural enterprises to manage consolidated land; and simultaneously developing specialized service organizations for machinery operation, unified pest control, processing, and sales to lower the barriers and risks of scale operation. Socialized services can compensate for labor shortages, ultimately achieving intensive, mechanized, and modern land management to enhance agricultural production efficiency and profitability. Crops cultivated on rural land should also be more varied. Expanding crop varieties can foster agricultural economic resilience, diversify economies, and expand crop sales channels.

Third, the nation ought to encourage the growth of smart agriculture. The growth of smart agriculture will lessen the influence of the rural population on the agricultural economy’s resilience and can, to some extent, lessen the reliance of agricultural economic resilience on the number of rural residents.

Nevertheless, certain limitations exist in this paper. First, due to constraints in data availability and model complexity considerations, climate change, price volatility, and subsidy policies were not incorporated into the current model. The omission of these important factors implies that the model may overestimate agricultural economic resilience and fails to capture major sources of vulnerability arising from price collapses or extreme climate events. Future research will focus on collecting relevant data and developing more sophisticated modeling frameworks to integrate these key drivers. Second, this study is constrained by its single-disciplinary perspective, making it difficult to explain the influence of sociological and ecological factors within bidirectional urban–rural population flow. Future research could introduce the Complex Adaptive Systems (CAS) theory, integrating knowledge from sociology and environmental science to quantify the contribution of social and ecological factors to bidirectional urban–rural population flow. Third, mobility itself is not a homogeneous phenomenon; its impacts and drivers may vary significantly by type (e.g., seasonal movement, forced displacement). Furthermore, broader production conditions, particularly agricultural mechanization, digitalization processes, and the precise design of agricultural development policies, are suggested by preliminary evidence in this study and existing literature as critical contextual variables. Future research should prioritize collecting granular data to enable the systematic differentiation and quantitative analysis of these types and specific conditions, thereby providing more targeted policy insights.